sklearn学习总结（超全面）

阳光穿透心脏的1/2处 2022-05-30 09:44 242阅读 0赞

#### 前言 ####

sklearn想必不用我多介绍了，一句话，**她是机器学习领域中最知名的python模块之一，若想要在机器学习领域有一番建树，必绕不开sklearn**

sklearn的官网链接[http://scikit-learn.org/stable/index.html\#][http_scikit-learn.org_stable_index.html]

首先，放上一张官网上的sklearn的结构图：  
![在这里插入图片描述][70]

### 目录 ###

### 1. 分类、回归 ###

### 2. 降维 ###

### 3. 模型评估与选择 ###

### 4. 数据预处理 ###

<table> 
 <thead> 
 <tr> 
 <th>大类</th> 
 <th>小类</th> 
 <th>适用问题</th> 
 <th>实现</th> 
 <th>说明</th> 
 </tr> 
 </thead> 
 <tbody> 
 <tr> 
 <td>分类、回归</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.1 <a href="http://scikit-learn.org/stable/modules/linear_model.html" rel="nofollow">广义线性模型</a></td> 
 <td>1.1.1 普通最小二乘法</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html#sklearn.linear_model.LinearRegression" rel="nofollow">sklearn.linear_model.LinearRegression</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>注：本节中所有的回归模型皆为线性回归模型</td> 
 <td>1.1.2 Ridge/岭回归</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Ridge.html#sklearn.linear_model.Ridge" rel="nofollow">sklearn.linear_model.Ridge</a></td> 
 <td>解决两类回归问题： 一是样本少于变量个数 二是变量间存在共线性</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.3 Lasso</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Lasso.html#sklearn.linear_model.Lasso" rel="nofollow">sklearn.linear_model.Lasso</a></td> 
 <td>适合特征较少的数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.4 Multi-task Lasso</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.MultiTaskLasso.html#sklearn.linear_model.MultiTaskLasso" rel="nofollow">sklearn.linear_model.MultiTaskLasso</a></td> 
 <td>y值不是一元的回归问题</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.5 Elastic Net</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.ElasticNet.html#sklearn.linear_model.ElasticNet" rel="nofollow">sklearn.linear_model.ElasticNet</a></td> 
 <td>结合了Ridge和Lasso</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.6 Multi-task Elastic Net</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.MultiTaskElasticNet.html#sklearn.linear_model.MultiTaskElasticNet" rel="nofollow">sklearn.linear_model.MultiTaskElasticNet</a></td> 
 <td>y值不是一元的回归问题</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.7 Least Angle Regression(LARS)</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Lars.html#sklearn.linear_model.Lars" rel="nofollow">sklearn.linear_model.Lars</a></td> 
 <td>适合高维数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.8 LARS Lasso</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LassoLars.html#sklearn.linear_model.LassoLars" rel="nofollow">sklearn.linear_model.LassoLars</a></td> 
 <td>(1)适合高维数据使用 (2)LARS算法实现的lasso模型</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.9 Orthogonal Matching Pursuit (OMP)</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.OrthogonalMatchingPursuit.html#sklearn.linear_model.OrthogonalMatchingPursuit" rel="nofollow">sklearn.linear_model.OrthogonalMatchingPursuit</a></td> 
 <td>基于贪心算法实现</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.10 贝叶斯回归</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.BayesianRidge.html#sklearn.linear_model.BayesianRidge" rel="nofollow">sklearn.linear_model.BayesianRidge</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.ARDRegression.html#sklearn.linear_model.ARDRegression" rel="nofollow">sklearn.linear_model.ARDRegression</a></td> 
 <td>优点： (1)适用于手边数据(2)可用于在估计过程中包含正规化参数 缺点：耗时</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.11 Logistic regression</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegression" rel="nofollow">sklearn.linear_model.LogisticRegression</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.12 SGD(随机梯度下降法)</td> 
 <td>分类 /回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDClassifier.html#sklearn.linear_model.SGDClassifier" rel="nofollow">sklearn.linear_model.SGDClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDRegressor.html#sklearn.linear_model.SGDRegressor" rel="nofollow">sklearn.linear_model.SGDRegressor</a></td> 
 <td>适用于大规模数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.13 Perceptron</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Perceptron.html#sklearn.linear_model.Perceptron" rel="nofollow">sklearn.linear_model.Perceptron</a></td> 
 <td>适用于大规模数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.14 Passive Aggressive Algorithms</td> 
 <td>分类 /回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.PassiveAggressiveClassifier.html#sklearn.linear_model.PassiveAggressiveClassifier" rel="nofollow">sklearn.linear_model. PassiveAggressiveClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.PassiveAggressiveRegressor.html#sklearn.linear_model.PassiveAggressiveRegressor" rel="nofollow">sklearn.linear_model. PassiveAggressiveRegressor</a></td> 
 <td>适用于大规模数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.15 Huber Regression</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.HuberRegressor.html#sklearn.linear_model.HuberRegressor" rel="nofollow">sklearn.linear_model.HuberRegressor</a></td> 
 <td>能够处理数据中有异常值的情况</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.1.16 多项式回归</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.PolynomialFeatures.html#sklearn.preprocessing.PolynomialFeatures" rel="nofollow">sklearn.preprocessing.PolynomialFeatures</a></td> 
 <td>通过PolynomialFeatures将非线性特征转化成多项式形式，再用线性模型进行处理</td> 
 </tr> 
 <tr> 
 <td>1.2 <a href="http://scikit-learn.org/stable/modules/lda_qda.html" rel="nofollow">线性和二次判别分析</a></td> 
 <td>1.2.1 LDA</td> 
 <td>分类/降维</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.discriminant_analysis.LinearDiscriminantAnalysis.html#sklearn.discriminant_analysis.LinearDiscriminantAnalysis" rel="nofollow">sklearn.discriminant_analysis. LinearDiscriminantAnalysis</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.2.2 QDA</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis.html#sklearn.discriminant_analysis.QuadraticDiscriminantAnalysis" rel="nofollow">sklearn.discriminant_analysis. QuadraticDiscriminantAnalysis</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.3 <a href="http://scikit-learn.org/stable/modules/kernel_ridge.html" rel="nofollow">核岭回归</a></td> 
 <td>简称KRR</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.kernel_ridge.KernelRidge.html#sklearn.kernel_ridge.KernelRidge" rel="nofollow">sklearn.kernel_ridge.KernelRidge</a></td> 
 <td>将核技巧应用到岭回归(1.1.2)中，以实现非线性回归</td> 
 </tr> 
 <tr> 
 <td>1.4 <a href="http://scikit-learn.org/stable/modules/svm.html" rel="nofollow">支持向量机</a></td> 
 <td>1.4.1 SVC,NuSVC,LinearSVC</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html#sklearn.svm.SVC" rel="nofollow">sklearn.svm.SVC</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.NuSVC.html#sklearn.svm.NuSVC" rel="nofollow">sklearn.svm.NuSVC</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.LinearSVC.html#sklearn.svm.LinearSVC" rel="nofollow">sklearn.svm.LinearSVC</a></td> 
 <td>SVC可用于非线性分类，可指定核函数； NuSVC与SVC唯一的不同是可控制支持向量的个数; LinearSVC用于线性分类</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.4.2 SVR,NuSVR,LinearSVR</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.SVR.html#sklearn.svm.SVR" rel="nofollow">sklearn.svm.SVR</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.NuSVC.html#sklearn.svm.NuSVR" rel="nofollow">sklearn.svm.NuSVR</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.LinearSVC.html#sklearn.svm.LinearSVR" rel="nofollow">sklearn.svm.LinearSVR</a></td> 
 <td>同上，将"分类"变成"回归"即可</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.4.3 OneClassSVM</td> 
 <td>异常检测</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.svm.OneClassSVM.html#sklearn.svm.OneClassSVM" rel="nofollow">sklearn.svm.OneClassSVM</a></td> 
 <td>无监督实现异常值检测</td> 
 </tr> 
 <tr> 
 <td>1.5 <a href="http://scikit-learn.org/stable/modules/sgd.html" rel="nofollow">随机梯度下降</a></td> 
 <td>同1.1.12</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.6 <a href="http://scikit-learn.org/stable/modules/neighbors.html" rel="nofollow">最近邻</a></td> 
 <td>1.6.1 Unsupervised Nearest Neighbors</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.NearestNeighbors.html#sklearn.neighbors.NearestNeighbors" rel="nofollow">sklearn.neighbors.NearestNeighbors</a></td> 
 <td>无监督实现K近邻的寻找</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.6.2 Nearest Neighbors Classification</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html#sklearn.neighbors.KNeighborsClassifier" rel="nofollow">sklearn.neighbors.KNeighborsClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.RadiusNeighborsClassifier.html#sklearn.neighbors.RadiusNeighborsClassifier" rel="nofollow">sklearn.neighbors.RadiusNeighborsClassifier</a></td> 
 <td>(1)不太适用于高维数据 (2)两种实现只是距离度量不一样，后者更适合非均匀的采样</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.6.3 Nearest Neighbors Regression</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsRegressor.html#sklearn.neighbors.KNeighborsRegressor" rel="nofollow">sklearn.neighbors.KNeighborsRegressor</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.RadiusNeighborsRegressor.html#sklearn.neighbors.RadiusNeighborsRegressor" rel="nofollow">sklearn.neighbors.RadiusNeighborsRegressor</a></td> 
 <td>同上</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.6.5 Nearest Centroid Classifier</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.NearestCentroid.html#sklearn.neighbors.NearestCentroid" rel="nofollow">sklearn.neighbors.NearestCentroid</a></td> 
 <td>每个类对应一个质心，测试样本被分类到距离最近的质心所在的类别</td> 
 </tr> 
 <tr> 
 <td>1.7 <a href="http://scikit-learn.org/stable/modules/gaussian_process.html" rel="nofollow">高斯过程(GP/GPML)</a></td> 
 <td>1.7.1 GPR</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.gaussian_process.GaussianProcessRegressor.html#sklearn.gaussian_process.GaussianProcessRegressor" rel="nofollow">sklearn.gaussian_process. GaussianProcessRegressor</a></td> 
 <td>与KRR一样使用了核技巧</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.7.3 GPC</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.gaussian_process.GaussianProcessClassifier.html#sklearn.gaussian_process.GaussianProcessClassifier" rel="nofollow">sklearn.gaussian_process. GaussianProcessClassifier</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.8 <a href="http://scikit-learn.org/stable/modules/cross_decomposition.html" rel="nofollow">交叉分解</a></td> 
 <td>实现算法：CCA和PLS</td> 
 <td>–</td> 
 <td>–</td> 
 <td>用来计算两个多元数据集的线性关系，当预测数据比观测数据有更多的变量时，用PLS更好</td> 
 </tr> 
 <tr> 
 <td>1.9 <a href="http://scikit-learn.org/stable/modules/naive_bayes.html" rel="nofollow">朴素贝叶斯</a></td> 
 <td>1.9.1 高斯朴素贝叶斯</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.GaussianNB.html#sklearn.naive_bayes.GaussianNB" rel="nofollow">sklearn.naive_bayes.GaussianNB</a></td> 
 <td>处理特征是连续型变量的情况</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.9.2 多项式朴素贝叶斯</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.MultinomialNB.html#sklearn.naive_bayes.MultinomialNB" rel="nofollow">sklearn.naive_bayes.MultinomialNB</a></td> 
 <td>最常见，要求特征是离散数据</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.9.3 伯努利朴素贝叶斯</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.BernoulliNB.html#sklearn.naive_bayes.BernoulliNB" rel="nofollow">sklearn.naive_bayes.BernoulliNB</a></td> 
 <td>要求特征是离散的，且为布尔类型，即true和false，或者1和0</td> 
 </tr> 
 <tr> 
 <td>1.10 <a href="http://scikit-learn.org/stable/modules/tree.html" rel="nofollow">决策树</a></td> 
 <td>1.10.1 Classification</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html#sklearn.tree.DecisionTreeClassifier" rel="nofollow">sklearn.tree.DecisionTreeClassifier</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.10.2 Regression</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeRegressor.html#sklearn.tree.DecisionTreeRegressor" rel="nofollow">sklearn.tree.DecisionTreeRegressor</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.11 <a href="http://scikit-learn.org/stable/modules/ensemble.html" rel="nofollow">集成方法</a></td> 
 <td>1.11.1 Bagging</td> 
 <td>分类/回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.BaggingClassifier.html#sklearn.ensemble.BaggingClassifier" rel="nofollow">sklearn.ensemble.BaggingClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.BaggingRegressor.html#sklearn.ensemble.BaggingRegressor" rel="nofollow">sklearn.ensemble.BaggingRegressor</a></td> 
 <td>可以指定基学习器，默认为决策树</td> 
 </tr> 
 <tr> 
 <td>注：1和2属于集成方法中的并行化方法，3和4属于序列化方法</td> 
 <td>1.11.2 Forests of randomized trees</td> 
 <td>分类/回归</td> 
 <td>RandomForest(RF,随机森林): <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html#sklearn.ensemble.RandomForestClassifier" rel="nofollow">sklearn.ensemble.RandomForestClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html#sklearn.ensemble.RandomForestRegressor" rel="nofollow">sklearn.ensemble.RandomForestRegressor</a> ExtraTrees(RF改进): <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.ExtraTreesClassifier.html#sklearn.ensemble.ExtraTreesClassifier" rel="nofollow">sklearn.ensemble.ExtraTreesClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.ExtraTreesRegressor.html#sklearn.ensemble.ExtraTreesRegressor" rel="nofollow">sklearn.ensemble.ExtraTreesRegressor</a></td> 
 <td>基学习器为决策树</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.11.3 AdaBoost</td> 
 <td>分类/回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html#sklearn.ensemble.AdaBoostClassifier" rel="nofollow">sklearn.ensemble.AdaBoostClassifier</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostRegressor.html#sklearn.ensemble.AdaBoostRegressor" rel="nofollow">sklearn.ensemble.AdaBoostRegressor</a></td> 
 <td>可以指定基学习器，默认为决策树</td> 
 </tr> 
 <tr> 
 <td>号外：最近特别火的两个梯度提升算法，LightGBM和XGBoost (XGBoost提供了sklearn接口)</td> 
 <td>1.11.4 Gradient Tree Boosting</td> 
 <td>分类/回归</td> 
 <td>GBDT: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html#sklearn.ensemble.GradientBoostingClassifier" rel="nofollow">sklearn.ensemble.GradientBoostingClassifier</a> GBRT: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingRegressor.html#sklearn.ensemble.GradientBoostingRegressor" rel="nofollow">sklearn.ensemble.GradientBoostingRegressor</a></td> 
 <td>基学习器为决策树</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.11.5 Voting Classifier</td> 
 <td>分类</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.VotingClassifier.html#sklearn.ensemble.VotingClassifier" rel="nofollow">sklearn.ensemble.VotingClassifier</a></td> 
 <td>须指定基学习器</td> 
 </tr> 
 <tr> 
 <td>1.12 <a href="http://scikit-learn.org/stable/modules/multiclass.html" rel="nofollow">多类与多标签算法</a></td> 
 <td>–</td> 
 <td>–</td> 
 <td>–</td> 
 <td>sklearn中的分类算法都默认支持多类分类，其中LinearSVC、 LogisticRegression和GaussianProcessClassifier在进行多类分类时需指定参数multi_class</td> 
 </tr> 
 <tr> 
 <td>1.13 <a href="http://scikit-learn.org/stable/modules/feature_selection.html" rel="nofollow">特征选择</a></td> 
 <td>1.13.1 过滤法之方差选择法</td> 
 <td>特征选择</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.VarianceThreshold.html#sklearn.feature_selection.VarianceThreshold" rel="nofollow">sklearn.feature_selection.VarianceThreshold</a></td> 
 <td>特征选择方法分为3种：过滤法、包裹法和嵌入法。过滤法不用考虑后续学习器</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.13.2 过滤法之卡方检验</td> 
 <td>特征选择</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectKBest.html#sklearn.feature_selection.SelectKBest" rel="nofollow">sklearn.feature_selection.SelectKBest</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.13.3 包裹法之递归特征消除法</td> 
 <td>特征选择</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html#sklearn.feature_selection.RFE" rel="nofollow">sklearn.feature_selection.RFE</a></td> 
 <td>包裹法需考虑后续学习器，参数中需输入基学习器</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>1.13.4 嵌入法</td> 
 <td>特征选择</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.SelectFromModel.html#sklearn.feature_selection.SelectFromModel" rel="nofollow">sklearn.feature_selection.SelectFromModel</a></td> 
 <td>嵌入法是过滤法和嵌入法的结合，参数中也需输入基学习器</td> 
 </tr> 
 <tr> 
 <td>1.14 <a href="http://scikit-learn.org/stable/modules/label_propagation.html" rel="nofollow">半监督</a></td> 
 <td>1.14.1 Label Propagation</td> 
 <td>分类/回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.semi_supervised.LabelPropagation.html#sklearn.semi_supervised.LabelPropagation" rel="nofollow">sklearn.semi_supervised.LabelPropagation</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.semi_supervised.LabelSpreading.html#sklearn.semi_supervised.LabelSpreading" rel="nofollow">sklearn.semi_supervised.LabelSpreading</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.15 <a href="http://scikit-learn.org/stable/modules/isotonic.html" rel="nofollow">保序回归</a></td> 
 <td>–</td> 
 <td>回归</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.isotonic.IsotonicRegression.html#sklearn.isotonic.IsotonicRegression" rel="nofollow">sklearn.isotonic.IsotonicRegression</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>1.16 <a href="http://scikit-learn.org/stable/modules/calibration.html" rel="nofollow">概率校准</a></td> 
 <td>–</td> 
 <td>–</td> 
 <td>–</td> 
 <td>在执行分类时，获得预测的标签的概率</td> 
 </tr> 
 <tr> 
 <td>1.17 <a href="http://scikit-learn.org/stable/modules/neural_networks_supervised.html" rel="nofollow">神经网络模型</a></td> 
 <td>（待写）</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>降维</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>2.5 <a href="http://scikit-learn.org/stable/modules/decomposition.html#decompositions" rel="nofollow">降维</a></td> 
 <td>2.5.1 主成分分析</td> 
 <td>降维</td> 
 <td>PCA: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html#sklearn.decomposition.PCA" rel="nofollow">sklearn.decomposition.PCA</a> IPCA: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.IncrementalPCA.html#sklearn.decomposition.IncrementalPCA" rel="nofollow">sklearn.decomposition.IncrementalPCA</a> KPCA: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.KernelPCA.html#sklearn.decomposition.KernelPCA" rel="nofollow">sklearn.decomposition.KernelPCA</a> SPCA: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.SparsePCA.html#sklearn.decomposition.SparsePCA" rel="nofollow">sklearn.decomposition.SparsePCA</a></td> 
 <td>(1)IPCA比PCA有更好的内存效率，适合超大规模降维。 (2)KPCA可以进行非线性降维 (3)SPCA是PCA的变体，降维后返回最佳的稀疏矩阵</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>2.5.2 截断奇异值分解</td> 
 <td>降维</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.TruncatedSVD.html#sklearn.decomposition.TruncatedSVD" rel="nofollow">sklearn.decomposition.TruncatedSVD</a></td> 
 <td>可以直接对scipy.sparse矩阵处理</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>2.5.3 字典学习</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.SparseCoder.html#sklearn.decomposition.SparseCoder" rel="nofollow">sklearn.decomposition.SparseCoder</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.DictionaryLearning.html#sklearn.decomposition.DictionaryLearning" rel="nofollow">sklearn.decomposition.DictionaryLearning</a></td> 
 <td>SparseCoder实现稀疏编码，DictionaryLearning实现字典学习</td> 
 </tr> 
 <tr> 
 <td>模型评估与选择</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>3.1 <a href="http://scikit-learn.org/stable/modules/cross_validation.html" rel="nofollow">交叉验证/CV</a></td> 
 <td>3.1.1 分割训练集和测试集</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html#sklearn.model_selection.train_test_split" rel="nofollow">sklearn.model_selection.train_test_split</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.1.2 通过交叉验证评估score</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.cross_val_score.html#sklearn.model_selection.cross_val_score" rel="nofollow">sklearn.model_selection.cross_val_score</a></td> 
 <td>score对应性能度量，分类问题默认为accuracy_score，回归问题默认为r2_score</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.1.3 留一法LOO</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.LeaveOneOut.html#sklearn.model_selection.LeaveOneOut" rel="nofollow">sklearn.model_selection.LeaveOneOut</a></td> 
 <td>CV的特例</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.1.4 留P法LPO</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.LeavePOut.html#sklearn.model_selection.LeavePOut" rel="nofollow">sklearn.model_selection.LeavePOut</a></td> 
 <td>CV的特例</td> 
 </tr> 
 <tr> 
 <td>3.2 <a href="http://scikit-learn.org/stable/modules/grid_search.html" rel="nofollow">调参</a></td> 
 <td>3.2.1 网格搜索</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV" rel="nofollow">sklearn.model_selection.GridSearchCV</a></td> 
 <td>最常用的调参方法。可传入学习器、学习器参数范围、性能度量score(默认为accuracy_score或r2_score )等</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.2.2 随机搜索</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.RandomizedSearchCV.html#sklearn.model_selection.RandomizedSearchCV" rel="nofollow">sklearn.model_selection.RandomizedSearchCV</a></td> 
 <td>参数传入同上</td> 
 </tr> 
 <tr> 
 <td>3.3 <a href="http://scikit-learn.org/stable/modules/model_evaluation.html" rel="nofollow">性能度量</a></td> 
 <td>3.3.1 分类度量</td> 
 <td>–</td> 
 <td>–</td> 
 <td>对应交叉验证和调参中的score</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.3.2 回归度量</td> 
 <td>–</td> 
 <td>–</td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.3.3 聚类度量</td> 
 <td>–</td> 
 <td>–</td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>3.4 <a href="http://scikit-learn.org/stable/modules/model_persistence.html" rel="nofollow">模型持久性</a></td> 
 <td>–</td> 
 <td>–</td> 
 <td>–</td> 
 <td>使用pickle存放模型，可以使模型不用重复训练</td> 
 </tr> 
 <tr> 
 <td>3.5 <a href="http://scikit-learn.org/stable/modules/learning_curve.html#learning-curve" rel="nofollow">验证曲线</a></td> 
 <td>3.5.1 验证曲线</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.validation_curve.html#sklearn.model_selection.validation_curve" rel="nofollow">sklearn.model_selection.validation_curve</a></td> 
 <td>横轴为某个参数的值，纵轴为模型得分</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>3.5.2 学习曲线</td> 
 <td>–</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.learning_curve.html#sklearn.model_selection.learning_curve" rel="nofollow">sklearn.model_selection.learning_curve</a></td> 
 <td>横轴为训练数据大小，纵轴为模型得分</td> 
 </tr> 
 <tr> 
 <td>数据预处理</td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td>4.3 <a href="http://scikit-learn.org/stable/modules/preprocessing.html#preprocessing" rel="nofollow">数据预处理</a></td> 
 <td>4.3.1 标准化</td> 
 <td>数据预处理</td> 
 <td>标准化: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.scale.html#sklearn.preprocessing.scale" rel="nofollow">sklearn.preprocessing.scale</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html#sklearn.preprocessing.StandardScaler" rel="nofollow">sklearn.preprocessing.StandardScaler</a></td> 
 <td>scale与StandardScaler都是将将特征转化成标准正态分布(即均值为0，方差为1),且都可以处理scipy.sparse矩阵，但一般选择后者</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td></td> 
 <td>数据预处理</td> 
 <td> 区间缩放: <a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html#sklearn.preprocessing.MinMaxScaler" rel="nofollow">sklearn.preprocessing.MinMaxScaler</a> <a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MaxAbsScaler.html#sklearn.preprocessing.MaxAbsScaler" rel="nofollow">sklearn.preprocessing.MaxAbsScale</a></td> 
 <td>MinMaxScaler默认为0-1缩放，MaxAbsScaler可以处理scipy.sparse矩阵</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.2 非线性转换</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.QuantileTransformer.html#sklearn.preprocessing.QuantileTransformer" rel="nofollow">sklearn.preprocessing.QuantileTransformer</a></td> 
 <td>可以更少的受异常值的影响</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.3 归一化</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.Normalizer.html#sklearn.preprocessing.Normalizer" rel="nofollow">sklearn.preprocessing.Normalizer</a></td> 
 <td>将行向量转换为单位向量，目的在于样本向量在点乘运算或其他核函数计算相似性时，拥有统一的标准</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.4 二值化</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.Binarizer.html#sklearn.preprocessing.Binarizer" rel="nofollow">sklearn.preprocessing.Binarizer</a></td> 
 <td>通过设置阈值对定量特征处理，获取布尔值</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.5 哑编码</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html#sklearn.preprocessing.OneHotEncoder" rel="nofollow">sklearn.preprocessing.OneHotEncoder</a></td> 
 <td>对定性特征编码。也可用pandas.get_dummies实现</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.6 缺失值计算</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.Imputer.html#sklearn.preprocessing.Imputer" rel="nofollow">sklearn.preprocessing.Imputer</a></td> 
 <td>可用三种方式填充缺失值，均值（默认）、中位数和众数。也可用pandas.fillna实现</td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.7 多项式转换</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.PolynomialFeatures.html#sklearn.preprocessing.PolynomialFeatures" rel="nofollow">sklearn.preprocessing.PolynomialFeatures</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td>4.3.8 自定义转换</td> 
 <td>数据预处理</td> 
 <td><a href="http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.FunctionTransformer.html#sklearn.preprocessing.FunctionTransformer" rel="nofollow">sklearn.preprocessing.FunctionTransformer</a></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 <tr> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 <td></td> 
 </tr> 
 </tbody> 
</table>

我的[github][]上也有，欢迎大家去看

[http_scikit-learn.org_stable_index.html]: http://scikit-learn.org/stable/index.html#
[70]: /images/20220530/f55833b05b2e40979f138526d8d7922c.png
[github]: https://github.com/fuqiuai/learning_notes/blob/master/sklearn%E5%AD%A6%E4%B9%A0.md