Python机器学习笔记：XgBoost算法

素颜马尾好姑娘i 2022-02-12 23:21 605阅读 0赞

## 前言 ##

### 1，Xgboost简介 ###

Xgboost是Boosting算法的其中一种，Boosting算法的思想是将许多弱分类器集成在一起，形成一个强分类器。因为Xgboost是一种提升树模型，所以它是将许多树模型集成在一起，形成一个很强的分类器。而所用到的树模型则是CART回归树模型。

Xgboost是在GBDT的基础上进行改进，使之更强大，适用于更大范围。

Xgboost一般和sklearn一起使用，但是由于sklearn中没有集成Xgboost，所以才需要单独下载安装。

### 2，Xgboost的优点 ###

Xgboost算法可以给预测模型带来能力的提升。当我们对其表现有更多了解的时候，我们会发现他有如下优势：

#### 2.1  正则化 ####

实际上，Xgboost是以“正则化提升（regularized boosting）” 技术而闻名。Xgboost在代价函数里加入了正则项，用于控制模型的复杂度。正则项里包含了树的叶子节点个数，每个叶子节点上输出的score的L2模的平方和。从Bias-variance tradeoff角度来讲，正则项降低了模型的variance，使学习出来的模型更加简单，防止过拟合，这也是Xgboost优于传统GBDT的一个特征

#### 2.2  并行处理 ####

Xgboost工具支持并行。众所周知，Boosting算法是顺序处理的，也是说Boosting不是一种串行的结构吗？怎么并行的？注意Xgboost的并行不是tree粒度的并行。Xgboost也是一次迭代完才能进行下一次迭代的（第t次迭代的代价函数里包含）。Xgboost的并行式在特征粒度上的，也就是说每一颗树的构造都依赖于前一颗树。

我们知道，决策树的学习最耗时的一个步骤就是对特征的值进行排序（因为要确定最佳分割点），Xgboost在训练之前，预先对数据进行了排序，然后保存为block结构，后面的迭代中重复使用这个结构，大大减小计算量。这个block结构也使得并行成为了可能，在进行节点的分类时，需要计算每个特征的增益，大大减少计算量。这个block结构也使得并行成为了可能，在进行节点的分裂的时候，需要计算每个特征的增益，最终选增益最大的那个特征去做分裂，那么各个特征的增益计算就可以开多线程进行。

#### 2.3  灵活性 ####

Xgboost支持用户**自定义目标函数和评估函数**，只要目标函数二阶可导就行。它对模型增加了一个全新的维度，所以我们的处理不会受到任何限制。

#### 2.4  缺失值处理 ####

对于特征的值有缺失的样本，Xgboost可以自动学习出他的分裂方向。Xgboost内置处理缺失值的规则。用户需要提供一个和其他样本不同的值，然后把它作为一个参数穿进去，以此来作为缺失值的取值。Xgboost在不同节点遇到缺失值时采用不同的处理方法，并且会学习未来遇到缺失值时的处理方法。

#### 2.5  剪枝 ####

Xgboost先从顶到底建立所有可以建立的子树，再从底到顶反向机芯剪枝，比起GBM，这样不容易陷入局部最优解

#### 2.6  内置交叉验证 ####

Xgboost允许在每一轮Boosting迭代中使用交叉验证。因此可以方便的获得最优Boosting迭代次数，而GBM使用网格搜索，只能检测有限个值。

### 3，Xgboost的离线安装 ###

1，[点击此处][Link 1]，下载对应自己Python版本的网址。

![1226410-20190129101052219-1623839238.png][]

2，输入安装的程式：

![1226410-20190129101537407-472954685.png][]

## Xgboost模型详解 ##

### 1，Xgboost能加载的各种数据格式解析 ###

Xgboost可以加载多种数据格式的训练数据：

下面一一列举：

记载libsvm格式的数据

记载二进制的缓存文件

加载numpy的数组

将scipy.sparse格式的数据转化为Dmatrix格式

将Dmatrix格式的数据保存成Xgboost的二进制格式，在下次加载时可以提高加载速度，使用方法如下：

可以使用如下方式处理DMatrix中的缺失值

当需要给样本设置权重时，可以用如下方式：

### 2，Xgboost的模型参数 ###

Xgboost使用key-value字典的方式存储参数

在运行Xgboost之前，必须设置三种类型成熟：general parameters，booster parameters和task parameters：

**通用参数（General Parameters）**：该参数控制在提升（boosting）过程中使用哪种booster，常用的booster有树模型（tree）和线性模型（linear model）

**Booster参数（Booster Parameters）**：这取决于使用哪种booster

**学习目标参数（Task Parameters）**：控制学习的场景，例如在回归问题中会使用不同的参数控制排序

#### 2.1， 通用参数 ####

booster\[default = gbtree\]：有两种模型可以选择gbtree 和 gblinear。gbtree使用基于树的模型进行提升计算，gblinear使用线性模型进行提升计算。缺省值为gbtree。

silent\[default = 0\] ：取0时表示打印出运行时信息，取1时表示已默认方式运行，不打印运行信息。缺省值为0

verbosity：\[默认 =1\] 打印消息的详细程度，有效值为0（静默），1（警告），2（信息），3（调试）。有时，XGBoost会尝试根据启发式更改配置，启动式显示为警告消息。如果出现意外，请尝试增加详细程度。

nthread：Xgboost运行的线程数，缺省值是当前系统可以获得的最大线程数

num\_pbuffer：预测缓冲区大小，通常设置为训练实例的数目。缓冲用于保存最后一步提升的预测结果，无需人为设置。

num\_feature：Boosting过程中用到的特征维数，设置为特征个数，Xgboost会自动设置，无需认为设置

disable\_default\_eval\_metric ：\[默认 = 0\] 标记以禁止默认度量标准，设置为> 0 以禁止。

#### 2.2，tree booster参数 ####

eta \[default=0.3\] ：为了防止过拟合，更新过程中用到的收缩步长。在每次提升计算之后，算法会直接获得新特征的权重。 eta通过缩减特征的权重使提升计算过程更加保守。缺省值为0.3   
取值范围为：\[0,1\]。典型值为0.01-0.2。

gamma \[default=0\] ：在节点分裂时，只有分裂后损失函数的值下降了，才会分裂这个节点。Gamma指定了节点分裂所需的最小损失函数下降值。这个参数的值越大，算法越保守。这个参数的值和损失函数息息相关，所以是需要调整的。取值范围为：\[0,∞\]

max\_depth \[default=6\] ：数的最大深度。缺省值为6 。取值范围为：\[1,∞\]。需要使用CV函数来进行调优。典型值：3-10

min\_child\_weight \[default=1\] ：孩子节点中最小的样本权重和。如果一个叶子节点的样本权重和小于min\_child\_weight则拆分过程结束。在现行回归模型中，这个参数是指建立每个模型所需要的最小样本数。这个参数用于避免过拟合。当它的值较大时，可以避免模型学习到局部的特殊样本。   
但是如果这个值过高，会导致欠拟合。这个参数需要使用CV来调整。取值范围为：\[0,∞\]

max\_delta\_step \[default=0\] ：我们允许每个树的权重被估计的值。如果它的值被设置为0，意味着没有约束；如果它被设置为一个正值，它能够使得更新的步骤更加保守。通常这个参数是没有必要的，但是如果在逻辑回归中类极其不平衡这时候他有可能会起到帮助作用。把它范围设置为1-10之间也许能控制更新。 取值范围为：\[0,∞\]

subsample \[default=1\] ：用于训练模型的子样本占整个样本集合的比例。如果设置为0.5则意味着XGBoost将随机的从整个样本集合中随机的抽取出50%的子样本建立树模型，这能够防止过拟合。 取值范围为：(0,1\]

colsample\_bytree , colsample\_bylevel , colsample\_bynode \[默认 =1\] 这是对列的子采样参数家族，所有的colsample\_by\* 参数的范围均为（0,1），默认值为1.

colsample\_bytree \[default=1\] ：在建立树时对特征采样的比例。缺省值为1 。取值范围为：(0,1\]

colsample\_bylevel ：是每个级别的列的子采样率，对于树中达到的每个新深度级别，子采样都会发生一次，列是从当前树的列集中进行子采样的。

colsample\_bynode：是每个节点（拆分）的列的子采样率。每次评估新的拆分时，都会发生一次子采样。列是从为当前级别选择的列集中进行二次采样的。

colsample\_by\* ：参数累积起作用。例如，与64个功能的组合将在每个拆分中留下4个功能可供选择。

#### 2.3，Linear Booster参数 ####

lambda\[default = 1\]：L2正则的惩罚系数，增加此值可以使模型更加保守

alpha\[default = 1\]：L1正则的惩罚系数，增加此值可以使模型更加保守

lambda\_bias：在偏置上的L2正则。缺省值为0（在L1上没有偏置项的正则，因为L1偏置时不需要）

#### 2.4，学习目标参数 ####

这个参数是来控制理想的优化目标和每一步结果的度量方法。

objective\[default = reg:linear\] ：定义学习任务及相应的学习目标，可选择的目标函数如下：

*  reg:linear” —— 线性回归。
 *  “reg:logistic”—— 逻辑回归。
 *  “binary:logistic”—— 二分类的逻辑回归问题，输出为概率。
 *  “binary:logitraw”—— 二分类的逻辑回归问题，输出的结果为wTx。
 *  “count:poisson”—— 计数问题的poisson回归，输出结果为poisson分布。在poisson回归中，max\_delta\_step的缺省值为0.7。(used to safeguard optimization)
 *  “multi:softmax” –让XGBoost采用softmax目标函数处理多分类问题，同时需要设置参数num\_class（类别个数）
 *  “multi:softprob” –和softmax一样，但是输出的是ndata \* nclass的向量，可以将该向量reshape成ndata行nclass列的矩阵。没行数据表示样本所属于每个类别的概率。
 *  “rank:pairwise” –set XGBoost to do ranking task by minimizing the pairwise loss

base\_score\[default = 0.5\]

所有实例的初始化预测分数，全局偏置；当有足够的迭代次数时，改变这个值将不会有太大的影响

eval\_metric\[ default according to objective\]

校验数据所需要的评价指标，不同的目标函数将会有缺省的评价指标（rmse for regression and error for classification  mean average precision for ranking）

用户可以添加多种评估指标，对于Python用户要以list传递参数对给程序，而不是map参数list参数不会覆盖eval\_metric

可供的选择如下：

![1226410-20190129171411164-1583579393.png][]

seed\[default=0\] ：随机数的种子，缺省值为0

### 3，Xgboost基本方法和默认参数 ###

parms：这是一个字典，里面包含着训练中的参数关键字和对应的值，形式是parms = \{'booster':'gbtree','eta':0.1\}

dtrain：训练的数据

num\_boost\_round：这是指提升迭代的个数

evals：这是一个列表，用于对训练过程中进行评估列表中的元素。形式是evals = \[(dtrain,'train'),(dval,'val')\] 或者是 evals =\[(dtrain,'train')\] ，对于第一种情况，它使得我们可以在训练过程中观察验证集的效果。

obj ：自定义目的函数

feval：自定义评估函数

maximize：是否对评估函数进行最大化

early\_stopping\_rounds：早起停止次数，假设为100，验证集的误差迭代到一定程度在100次内不能再继续降低，就停止迭代。这要求evals里至少有一个元素，如果有多个，按照最后一个去执行。返回的是最后的迭代次数（不是最好的）。如果early\_stopping\_rounds存在，则模型会生成三个属性，bst.best\_score ,bst.best\_iteration和bst.best\_ntree\_limit

evals\_result：字典，存储在watchlist中的元素的评估结果

verbose\_eval（可以输入布尔型或者数值型）：也要求evals里至少有一个元素，如果为True，则对evals中元素的评估结果会输出在结果中；如果输入数字，假设为5，则每隔5个迭代输出一次。

learning\_rates：每一次提升的学习率的列表

xgb\_model：在训练之前用于加载的xgb\_model

### 4，模型训练 ###

有了参数列表和数据就可以训练模型了

### 5，模型预测 ###

### 6，保存模型 ###

在训练完成之后可以将模型保存下来，也可以查看模型内部的结构

导出模型和特征映射（Map）

你可以导出模型到txt文件并浏览模型的含义：

### 7，加载模型 ###

通过如下方式可以加载模型

## Xgboost实战 ##

Xgboost有两大类接口：Xgboost原生接口 和sklearn接口，并且Xgboost能够实现分类回归两种任务。下面对这四种情况做以解析。

###  1，基于Xgboost原生接口的分类 ###

<table style="border-collapse:collapse;border-spacing:0px;border:1px solid #C0C0C0;width:626px;background:none;line-height:1.1em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;font-size:12px;"> 
 <tbody style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
  <tr style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;width:35px;color:rgb(175,175,175);"> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      1 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      2 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      3 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      4 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      5 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      6 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      7 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      8 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      9 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      10 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      11 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      12 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      13 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      14 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      15 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      16 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      17 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      18 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      19 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      20 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      21 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      22 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      23 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      24 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      25 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      26 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      27 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      28 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      29 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      30 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      31 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      32 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      33 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      34 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      35 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      36 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      37 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      38 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      39 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      40 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      41 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      42 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      43 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      44 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      45 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      46 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      47 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      48 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      49 
    </div> </td> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;"> 
    <div style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.datasets import load_iris</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">import xgboost&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">as</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgb</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgboost import plot_importance</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">import matplotlib.pyplot&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">as</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plt</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.model_selection import train_test_split</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.metrics import accuracy_score&nbsp; # 准确率</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 记载样本数据集</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">iris = load_iris()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X,y = iris.data,iris.target</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 数据集分割</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.2,random_state=123457)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 算法参数</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">params</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">= { </code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'booster'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'gbtree'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'objective'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'multi:softmax'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'num_class'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:3,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'gamma'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.1,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'max_depth'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:6,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'lambda'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:2,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'subsample'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.7,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'colsample_bytree'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.7,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'min_child_weight'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:3,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'slient'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:1,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'eta'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.1,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'seed'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:1000,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'nthread'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:4,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">}</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plst =&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">params</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">.items()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 生成数据集格式</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">dtrain = xgb.DMatrix(X_train,y_train)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">num_rounds = 500</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># xgboost模型训练</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">model = xgb.train(plst,dtrain,num_rounds)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 对测试集进行预测</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">dtest = xgb.DMatrix(X_test)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y_pred = model.predict(dtest)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 计算准确率</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">accuracy = accuracy_score(y_test,y_pred)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">print(</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'accuarcy:%.2f%%'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">%(accuracy*100))</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 显示重要特征</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plot_importance(model)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plt.show()</code> 
     </div> 
    </div> </td> 
  </tr> 
 </tbody> 
</table>

输出预测正确率以及特征重要性：

![1226410-20190129114158285-1212995606.png][]

### 2，基于Xgboost原生接口的回归 ###

<table style="border-collapse:collapse;border-spacing:0px;border:1px solid #C0C0C0;width:617px;background:none;line-height:1.1em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;font-size:12px;"> 
 <tbody style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
  <tr style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;width:35px;color:rgb(175,175,175);"> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      1 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      2 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      3 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      4 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      5 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      6 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      7 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      8 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      9 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      10 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      11 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      12 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      13 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      14 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      15 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      16 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      17 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      18 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      19 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      20 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      21 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      22 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      23 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      24 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      25 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      26 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      27 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      28 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      29 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      30 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      31 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      32 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      33 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      34 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      35 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      36 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      37 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      38 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      39 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      40 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      41 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      42 
    </div> </td> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;"> 
    <div style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">import xgboost&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">as</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgb</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgboost import plot_importance</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">matplotlib import pyplot&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">as</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plt</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.model_selection import train_test_split</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.datasets import load_boston</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.metrics import mean_squared_error</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 加载数据集,此数据集时做回归的</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">boston = load_boston()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X,y = boston.data,boston.target</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># Xgboost训练过程</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.2,random_state=0)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 算法参数</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">params</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">= { </code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'booster'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'gbtree'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'objective'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'reg:gamma'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'gamma'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.1,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'max_depth'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:5,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'lambda'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:3,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'subsample'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.7,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'colsample_bytree'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.7,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'min_child_weight'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:3,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'slient'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:1,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'eta'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:0.1,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'seed'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:1000,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">'nthread'</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">:4,</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">}</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">dtrain = xgb.DMatrix(X_train,y_train)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">num_rounds = 300</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plst =&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">params</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">.items()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">model = xgb.train(plst,dtrain,num_rounds)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 对测试集进行预测</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">dtest = xgb.DMatrix(X_test)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">ans = model.predict(dtest)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 显示重要特征</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plot_importance(model)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">plt.show()</code> 
     </div> 
    </div> </td> 
  </tr> 
 </tbody> 
</table>

重要特征（值越大，说明该特征越重要）显示结果：

![1226410-20190129115848310-1170440550.png][]

###  3，Xgboost使用sklearn接口的分类（推荐） ###

#### XGBClassifier ####

<table style="border-collapse:collapse;border-spacing:0px;border:1px solid #C0C0C0;width:664px;background:none;line-height:1.1em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;font-size:12px;"> 
 <tbody style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
  <tr style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;width:35px;color:rgb(175,175,175);"> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      1 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      2 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      3 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      4 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      5 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      6 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      7 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      8 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      9 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      10 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      11 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      12 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      13 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      14 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      15 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      16 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      17 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      18 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      19 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      20 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      21 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      22 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      23 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      24 
    </div> </td> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;"> 
    <div style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgboost.sklearn import XGBClassifier</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">clf = XGBClassifier(</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">silent=0,&nbsp; # 设置成1则没有运行信息输出，最好是设置为0，是否在运行升级时打印消息</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># nthread = 4&nbsp; # CPU 线程数 默认最大</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">learning_rate=0.3 , # 如同学习率</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">min_child_weight = 1,</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 这个参数默认为1，是每个叶子里面h的和至少是多少，对正负样本不均衡时的0-1分类而言</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 假设h在0.01附近，min_child_weight为1 意味着叶子节点中最少需要包含100个样本</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># 这个参数非常影响结果，控制叶子节点中二阶导的和的最小值，该参数值越小，越容易过拟合</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">max_depth=6, # 构建树的深度，越大越容易过拟合</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">gamma = 0,# 树的叶子节点上做进一步分区所需的最小损失减少，越大越保守，一般0.1 0.2这样子</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">subsample=1, # 随机采样训练样本，训练实例的子采样比</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">max_delta_step=0,&nbsp; # 最大增量步长，我们允许每个树的权重估计</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">colsample_bytree=1, # 生成树时进行的列采样</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">reg_lambda=1, #控制模型复杂度的权重值的L2正则化项参数，参数越大，模型越不容易过拟合</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># reg_alpha=0, # L1正则项参数</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># scale_pos_weight =1 # 如果取值大于0的话，在类别样本不平衡的情况下有助于快速收敛，平衡正负权重</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># objective = 'multi:softmax', # 多分类问题，指定学习任务和响应的学习目标</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># num_class = 10,&nbsp; # 类别数，多分类与multisoftmax并用</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">n_estimators=100,&nbsp; # 树的个数</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">seed = 1000,&nbsp; # 随机种子</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># eval_metric ='auc'</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">)</code> 
     </div> 
    </div> </td> 
  </tr> 
 </tbody> 
</table>

#### 基于Sckit-learn接口的分类 ####

输出结果：

![1226410-20190129152059634-1248365127.png][]

### 4，基于Scikit-learn接口的回归 ###

![1226410-20190129152951940-323894783.png][]

## Xgboost参数调优的一般方法 ##

### 调参步骤： ###

1，选择较高的学习速率（learning rate）。一般情况下，学习速率的值为0.1.但是，对于不同的问题，理想的学习速率有时候会在0.05~0.3之间波动。选择对应于此学习速率的理想决策树数量。Xgboost有一个很有用的函数“cv”，这个函数可以在每一次迭代中使用交叉验证，并返回理想的决策树数量。

2，对于给定的学习速率和决策树数量，进行决策树特定参数调优（max\_depth , min\_child\_weight , gamma , subsample,colsample\_bytree）在确定一棵树的过程中，我们可以选择不同的参数。

3，Xgboost的正则化参数的调优。（lambda , alpha）。这些参数可以降低模型的复杂度，从而提高模型的表现。

4，降低学习速率，确定理想参数。

下面详细的进行这些操作。

### 第一步：确定学习速率和tree\_based参数调优的估计器数目 ###

为了确定Boosting参数，我们要先给其他参数一个初始值。咱们先按照如下方法取值：

*  1，max\_depth = 5：这个参数的取值最好在3-10之间，我选的起始值为5，但是你可以选择其他的值。起始值在4-6之间都是不错的选择。
 *  2，min\_child\_weight = 1 ：这里选择了一个比较小的值，因为这是一个极不平衡的分类问题。因此，某些叶子节点下的值会比较小。
 *  3，gamma = 0 :起始值也可以选择其它比较小的值，在0.1到0.2之间就可以，这个参数后继也是要调整的。
 *  4，subsample,colsample\_bytree = 0.8  这个是最常见的初始值了。典型值的范围在0.5-0.9之间。
 *  5，scale\_pos\_weight =1 这个值时因为类别十分不平衡。

注意，上面这些参数的值知识一个初始的估计值，后继需要调优。这里把学习速率就设成默认的0.1。然后用Xgboost中的cv函数来确定最佳的决策树数量。

### 第二步：max\_depth和min\_weight参数调优 ###

我们先对这两个参数调优，是因为他们对最终结果有很大的影响。首先，我们先大范围地粗略参数，然后再小范围的微调。

注意：在这一节我会进行高负荷的栅格搜索（grid search），这个过程大约需要15-30分钟甚至更久，具体取决于你系统的性能，你也可以根据自己系统的性能选择不同的值。

### 第三步：gamma参数调优 ###

在已经调整好其他参数的基础上，我们可以进行gamma参数的调优了。Gamma参数取值范围很大，这里我们设置为5，其实你也可以取更精确的gamma值。

### 第四步：调整subsample 和 colsample\_bytree参数 ###

尝试不同的subsample 和 colsample\_bytree 参数。我们分两个阶段来进行这个步骤。这两个步骤都取0.6,0.7,0.8,0.9作为起始值。

### 第五步：正则化参数调优 ###

由于gamma函数提供了一种更加有效的降低过拟合的方法，大部分人很少会用到这个参数，但是我们可以尝试用一下这个参数。

### 第六步：降低学习速率 ###

最后，我们使用较低的学习速率，以及使用更多的决策树，我们可以用Xgboost中CV函数来进行这一步工作。

总结一下，要想模型的表现有大幅的提升，调整每个参数带来的影响也必须清楚，仅仅靠着参数的调整和模型的小幅优化，想要让模型的表现有个大幅度提升是不可能的。要想模型的表现有质的飞跃，需要依靠其他的手段。诸如，特征工程(feature egineering) ，模型组合(ensemble of model),以及堆叠(stacking)等。

## XGBoost输出特征重要性以及筛选特征 ##

### 1，梯度提升算法是如何计算特征重要性的？ ###

使用梯度提升算法的好处是在提升树被创建后，可以相对直接地得到每个属性的重要性得分。一般来说，重要性分数，衡量了特征在模型中的提升决策树构建中的价值。一个属性越多的被用来在模型中构建决策树，它的重要性就相对越高。

属性重要性是通过对数据集中的每个属性进行计算，并进行排序得到。在单个决策树中通过每个属性分裂点改进性能度量的量来计算属性重要性。由节点负责加权和记录次数，也就是说一个属性对分裂点改进性能度量越大（越靠近根节点），权值越大；被越多提升树所选择，属性越重要。性能度量可以是选择分裂节点的Gini纯度，也可以是其他度量函数。

最终将一个属性在所有提升树中的结果进行加权求和后然后平均，得到重要性得分。

### 2，绘制特征重要性 ###

一个已训练的Xgboost模型能够自动计算特征重要性，这些重要性得分可以通过成员变量feature\_importances\_得到。可以通过如下命令打印：

我们可以直接在条形图上绘制这些分数，以便获得数据集中每个特征的相对重要性的直观显示，例如：

我们可以通过在the Pima Indians onset of diabetes 数据集上训练XGBoost模型来演示，并从计算的特征重要性中绘制条形图。

运行这个实例，首先输出特征重要性分数：

相对重要性条形图：

![1226410-20190227170129383-1069925183.png][]

这种绘制的缺点在于，只显示了特征重要性而没有排序，可以在绘制之前对特征重要性得分进行排序。

通过内建的绘制函数进行特征重要性得分排序后的绘制，这个函数就是plot\_importance()，示例如下：

示例得到条形图：

![1226410-20190227170428627-327092794.png][]

根据其在输入数组的索引，特征被自动命名为f0~f3，在问题描述中手动的将这些索引映射到名称，我们可以看到，f2具有最高的重要性，f1具有最低的重要性。

### 3，根据Xgboost特征重要性得分进行特征选择 ###

特征重要性得分，可以用于在scikit-learn中进行特征选择。通过SelectFromModel类实现，该类采用模型并将数据集转换为具有选定特征的子集。这个类可以采取预先训练的模型，例如在整个数据集上训练的模型。然后，它可以阈值来决定选择哪些特征。当在SelectFromModel实例上调用transform()方法时，该阈值被用于在训练集和测试集上一致性选择相同特征。

在下面的示例中，我们首先在训练集上训练xgboost模型，然后在测试上评估。使用从训练数据集计算的特征重要性，然后，将模型封装在一个SelectFromModel实例中。我们使用这个来选择训练集上的特征，用所选择的特征子集训练模型，然后在相同的特征方案下对测试集进行评估。

我们可以通过测试多个阈值，来从特征重要性中选择特征。具体而言，每个输入变量的特征重要性，本质上允许我们通过重要性来测试每个特征子集。

完整代码如下：

<table style="border-collapse:collapse;border-spacing:0px;border:1px solid #C0C0C0;background:none;line-height:1.1em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;font-size:12px;"> 
 <tbody style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
  <tr style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;width:35px;color:rgb(175,175,175);"> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      1 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      2 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      3 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      4 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      5 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      6 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      7 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      8 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      9 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      10 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      11 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      12 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      13 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      14 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      15 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      16 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      17 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      18 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      19 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      20 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      21 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      22 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      23 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      24 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      25 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      26 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      27 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      28 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      29 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      30 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      31 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      32 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      33 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      34 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      35 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      36 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      37 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      38 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      39 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      40 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      41 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      42 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      43 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      44 
    </div> 
    <div style="background:rgb(244,244,244) none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      45 
    </div> 
    <div style="background-image:none;border-width:0px 2px 0px 0px;border-right-style:solid;border-right-color:rgb(108,226,108);line-height:1.8em;text-align:right;vertical-align:baseline;">
      46 
    </div> </td> 
   <td style="border:1px solid #C0C0C0;border-collapse:collapse;background:none;line-height:1.1em;vertical-align:baseline;"> 
    <div style="background:none;border:0px;line-height:1.1em;vertical-align:baseline;"> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># plot feature importance manually</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">import numpy&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">as</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">np</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgboost import XGBClassifier</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">matplotlib import pyplot</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.datasets import load_iris</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">xgboost import plot_importance</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.model_selection import train_test_split</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.metrics import accuracy_score</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">from</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">sklearn.feature_selection import SelectFromModel</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># load data</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">dataset = load_iris()</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># split data into X and y</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X = dataset.data</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y = dataset.target</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># split data into train and test sets</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.33,random_state=7)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># fit model no training data</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">model = XGBClassifier()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">model.fit(X_train, y_train)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># feature importance</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">print(model.feature_importances_)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># make predictions for test data and evaluate</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y_pred = model.predict(X_test)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">predictions = [round(value)&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">for</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">value&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">in</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y_pred]</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">accuracy = accuracy_score(y_test,predictions)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">print(</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">"Accuracy:%.2f%%"</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">%(accuracy*100.0))</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;">
       &nbsp; 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;">#fit model using each importance as a threshold</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">thresholds = np.sort(model.feature_importances_)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">for</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">thresh&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">in</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">thresholds:</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># select features using threshold</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">selection = SelectFromModel(model,threshold=thresh,prefit=True )</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">select_X_train = selection.transform(X_train)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># train model</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">selection_model = XGBClassifier()</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">selection_model.fit(select_X_train, y_train)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#808080;"># eval model</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">select_X_test = selection.transform(X_test)</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y_pred = selection_model.predict(select_X_test)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">predictions = [round(value)&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">for</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">value&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:rgb(0,0,255);">in</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">y_pred]</code> 
     </div> 
     <div style="background:rgb(244,244,244) none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">accuracy = accuracy_score(y_test,predictions)</code> 
     </div> 
     <div style="background-image:none;border:0px;line-height:1.8em;vertical-align:baseline;"> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">&nbsp;&nbsp;&nbsp;&nbsp;</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">print(</code> 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;color:#0000FF;">"Thresh=%.3f, n=%d, Accuracy: %.2f%%"</code>&nbsp; 
      <code style="background:none;border:0px;line-height:1.8em;vertical-align:baseline;font-family:Consolas, 'Bitstream Vera Sans Mono', 'Courier New', Courier, monospace;">% (thresh, select_X_train.shape[1], accuracy * 100.0))</code> 
     </div> 
    </div> </td> 
  </tr> 
 </tbody> 
</table>

运行示例，得到输出：

我们可以看到，模型的性能通常随着所选择的特征的数量减少，在这一问题上，可以对测试集准确率和模型复杂度做一个权衡，例如选择三个特征，接受准确率为92%，这可能是对这样一个小数据集的清洗，但是对于更大的数据集和使用交叉验证作为模型评估方案可能是更有用的策略。

参考文献：

https://blog.csdn.net/waitingzby/article/details/81610495

https://blog.csdn.net/u011089523/article/details/72812019

https://blog.csdn.net/luanpeng825485697/article/details/79907149

https://xgboost.readthedocs.io/en/latest/parameter.html\#general-parameters

原文地址[https://www.cnblogs.com/wj-1314/p/9402324.html][https_www.cnblogs.com_wj-1314_p_9402324.html]

[Link 1]: https://www.lfd.uci.edu/~gohlke/pythonlibs/#xgboost
[1226410-20190129101052219-1623839238.png]: /images/20220213/12467f00e1bf4c92b9e29b0f9deed0cf.png
[1226410-20190129101537407-472954685.png]: /images/20220213/f2efc8fd8bd04a2e87fc315d7113ee15.png
[1226410-20190129171411164-1583579393.png]: /images/20220213/08614362c2534cde8457384ec329dbe6.png
[1226410-20190129114158285-1212995606.png]: /images/20220213/dd9c54612eb7497d9f247d722ce1fe0d.png
[1226410-20190129115848310-1170440550.png]: /images/20220213/76522c39d77c47489c6bb71565b6aad6.png
[1226410-20190129152059634-1248365127.png]: /images/20220213/b2bccb1f4ffd43c2bdd3d174a4b91d51.png
[1226410-20190129152951940-323894783.png]: /images/20220213/3775d6e039eb45d988057216f7007a66.png
[1226410-20190227170129383-1069925183.png]: /images/20220213/2bb235c810664b0d984476eb1267cb77.png
[1226410-20190227170428627-327092794.png]: /images/20220213/6b9ec072072a401fb14ba5d9112551bf.png
[https_www.cnblogs.com_wj-1314_p_9402324.html]: https://www.cnblogs.com/wj-1314/p/9402324.html