线性模型L1正则化——套索回归

from sklearn.model_selection import train_test_split
from sklearn.datasets import load_diabetes
X,y=load_diabetes().data,load_diabetes().target
X_train,X_test,y_train,y_test=train_test_split(X,y,random_state=8)

from sklearn.linear_model import Lasso
import numpy as np
lasso=Lasso().fit(X_train,y_train)
print("the coefficient:{}".format(lasso.coef_))
print('the intercept:{}'.format(lasso.intercept_))
print("the score of this model:{:.3f}".format(lasso.score(X_test,y_test)))
print("the model uses {}".format(np.sum(lasso.coef_!=0))+" features\n")

lasso01=Lasso(alpha=0.1,max_iter=100000).fit(X_train,y_train)
print("the coefficient:{}".format(lasso01.coef_))
print('the intercept:{}'.format(lasso01.intercept_))
print("the score of this model:{:.3f}".format(lasso01.score(X_test,y_test)))
print("the model uses {}".format(np.sum(lasso01.coef_!=0))+" features\n")

lasso001=Lasso(alpha=0.01,max_iter=100000).fit(X_train,y_train)
print("the coefficient:{}".format(lasso001.coef_))
print('the intercept:{}'.format(lasso001.intercept_))
print("the score of this model:{:.3f}".format(lasso001.score(X_test,y_test)))
print("the model uses {}".format(np.sum(lasso001.coef_!=0))+" features\n")

lasso0001=Lasso(alpha=0.001,max_iter=100000).fit(X_train,y_train)
print("the coefficient:{}".format(lasso0001.coef_))
print('the intercept:{}'.format(lasso0001.intercept_))
print("the score of this model:{:.3f}".format(lasso0001.score(X_test,y_test)))
print("the model uses {}".format(np.sum(lasso0001.coef_!=0))+" features\n")

lasso00001=Lasso(alpha=0.0001,max_iter=100000).fit(X_train,y_train)
print("the coefficient:{}".format(lasso00001.coef_))
print('the intercept:{}'.format(lasso00001.intercept_))
print("the score of this model:{:.3f}".format(lasso00001.score(X_test,y_test)))
print("the model uses {}".format(np.sum(lasso00001.coef_!=0))+" features\n")

import matplotlib.pyplot as plt
plt.plot(lasso.coef_,'s',label='Lasso alpha=1')
plt.plot(lasso01.coef_,'^',label='Lasso alpha=0.1')
plt.plot(lasso001.coef_,'v',label='Lasso alpha=0.01')
plt.plot(lasso0001.coef_,'o',label='Lasso alpha=0.001')
plt.plot(lasso00001.coef_,'*',label='Lasso alpha=0.0001')
plt.xlabel("coeffient index")
plt.ylabel("coeffient magnitude")
plt.legend(loc=(0,1.05))
plt.show()