【sklearn】【Generalized Linear Models】【1.1】

1. Generalized Linear Models

sklearn.linear_model	解释	样例
linear_model.AddRegression
linear_model.BayesianRidge
linear_model.ElasticNet
linear_model.ElasticNetCV
lnear_model.HuberRegressor
linear_model.Lars
linear_model.LarsCV
linear_model.Lasso	Lasso回归，可以将一些线性相关的变量系数变为0	from sklearn import linear_model reg = linear_model.Lasso(alpha=0.001) reg.fit([[0,0],[1,1],[2,2]], [0,1,2]) print reg.coef_, reg.intercept_ output: [ 0.9985 0. ] 0.0015
linear_model.LassoCV	Lasso回归，给定一个alpha列表，通关cross-validation选择合适的alpha	from sklearn import linear_model reg = linear_model.LassoCV(n_alphas = 3, alphas = [0.1, 1, 10]) reg.fit([[0,0],[1,1],[2,2]], [0,1,2]) print reg.coef_, reg.intercept_, reg.alpha_ output: [ 0.85 0. ] 0.15 0.1
linear_model.LassoLars
linear_model.LassoLarsCV
linear_model.LassoLarsIC
linear_model.LinearRegression	线性回归	from sklearn import linear_model reg = linear_model.LinearRegression() reg.fit([[0,0],[1,1],[2,2]], [0,1,2]) print reg.coef_, reg.intercept_ output: [ 0.5  0.5] 2.22044604925e-16
linear_model.LogsticRegression	逻辑回归，本质是一个分类算法	from sklearn import linear_model reg = linear_model.LogisticRegression() reg.fit([[0,1],[0,2],[0,3],[1,1],[1,2],[1,3]],[1,1,1,2,2,2]) print reg.coef_, reg.intercept_ print reg.predict([[0,100],[1,100],[1,4]]) output: [[ 1.04846721 -0.1433622 ]] [-0.13840227] [1 1 2]
linear_model.Ridge	岭回归，通过alpha控制系数的分散程度	from sklearn import linear_model reg = linear_model.Ridge(alpha=0.001) reg.fit([[0,0],[1,1],[2,2]], [0,1,2]) print reg.coef_, reg.intercept_ output: [ 0.49987503  0.49987503] 0.000249937515621
linear_model.RidgeCV	岭回归，给定一个alpha列表，通过cross-validation选择合适的alpha	from sklearn import linear_model reg = linear_model.RidgeCV([0.1, 1, 10]) reg.fit([[0,0],[1,1],[2,2]], [0,1,2]) print reg.coef_, reg.intercept_, reg.alpha_ output: [ 0.48780488  0.48780488] 0.0243902439024 0.1