python 机器学习之支持向量机非线性回归SVR模型

本文介绍了python 支持向量机非线性回归SVR模型，废话不多说，具体如下：

import numpy as np
import matplotlib.pyplot as plt

from sklearn import datasets, linear_model,svm
from sklearn.model_selection import train_test_split

def load_data_regression():
 '''
 加载用于回归问题的数据集
 '''
 diabetes = datasets.load_diabetes() #使用 scikit-learn 自带的一个糖尿病病人的数据集
 # 拆分成训练集和测试集，测试集大小为原始数据集大小的 1/4
 return train_test_split(diabetes.data,diabetes.target,test_size=0.25,random_state=0)

#支持向量机非线性回归SVR模型
def test_SVR_linear(*data):
 X_train,X_test,y_train,y_test=data
 regr=svm.SVR(kernel='linear')
 regr.fit(X_train,y_train)
 print('Coefficients:％s, intercept ％s'％(regr.coef_,regr.intercept_))
 print('Score: ％.2f' ％ regr.score(X_test, y_test))

# 生成用于回归问题的数据集
X_train,X_test,y_train,y_test=load_data_regression()
# 调用 test_LinearSVR
test_SVR_linear(X_train,X_test,y_train,y_test)

def test_SVR_poly(*data):
 '''
 测试 多项式核的 SVR 的预测性能随 degree、gamma、coef0 的影响.
 '''
 X_train,X_test,y_train,y_test=data
 fig=plt.figure()
 ### 测试 degree ####
 degrees=range(1,20)
 train_scores=[]
 test_scores=[]
 for degree in degrees:
   regr=svm.SVR(kernel='poly',degree=degree,coef0=1)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 ax=fig.add_subplot(1,3,1)
 ax.plot(degrees,train_scores,label="Training score ",marker='+' )
 ax.plot(degrees,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_poly_degree r=1")
 ax.set_xlabel("p")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1.)
 ax.legend(loc="best",framealpha=0.5)

### 测试 gamma，固定 degree为3， coef0 为 1 ####
 gammas=range(1,40)
 train_scores=[]
 test_scores=[]
 for gamma in gammas:
   regr=svm.SVR(kernel='poly',gamma=gamma,degree=3,coef0=1)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 ax=fig.add_subplot(1,3,2)
 ax.plot(gammas,train_scores,label="Training score ",marker='+' )
 ax.plot(gammas,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_poly_gamma r=1")
 ax.set_xlabel(r"$\gamma$")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1)
 ax.legend(loc="best",framealpha=0.5)
 ### 测试 r，固定 gamma 为 20，degree为 3 ######
 rs=range(0,20)
 train_scores=[]
 test_scores=[]
 for r in rs:
   regr=svm.SVR(kernel='poly',gamma=20,degree=3,coef0=r)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 ax=fig.add_subplot(1,3,3)
 ax.plot(rs,train_scores,label="Training score ",marker='+' )
 ax.plot(rs,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_poly_r gamma=20 degree=3")
 ax.set_xlabel(r"r")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1.)
 ax.legend(loc="best",framealpha=0.5)
 plt.show()

# 调用 test_SVR_poly
test_SVR_poly(X_train,X_test,y_train,y_test)

def test_SVR_rbf(*data):
 '''
 测试 高斯核的 SVR 的预测性能随 gamma 参数的影响
 '''
 X_train,X_test,y_train,y_test=data
 gammas=range(1,20)
 train_scores=[]
 test_scores=[]
 for gamma in gammas:
   regr=svm.SVR(kernel='rbf',gamma=gamma)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 fig=plt.figure()
 ax=fig.add_subplot(1,1,1)
 ax.plot(gammas,train_scores,label="Training score ",marker='+' )
 ax.plot(gammas,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_rbf")
 ax.set_xlabel(r"$\gamma$")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1)
 ax.legend(loc="best",framealpha=0.5)
 plt.show()

# 调用 test_SVR_rbf
test_SVR_rbf(X_train,X_test,y_train,y_test)

def test_SVR_sigmoid(*data):
 '''
 测试 sigmoid 核的 SVR 的预测性能随 gamma、coef0 的影响.
 '''
 X_train,X_test,y_train,y_test=data
 fig=plt.figure()

### 测试 gammam，固定 coef0 为 0.01 ####
 gammas=np.logspace(-1,3)
 train_scores=[]
 test_scores=[]

for gamma in gammas:
   regr=svm.SVR(kernel='sigmoid',gamma=gamma,coef0=0.01)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 ax=fig.add_subplot(1,2,1)
 ax.plot(gammas,train_scores,label="Training score ",marker='+' )
 ax.plot(gammas,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_sigmoid_gamma r=0.01")
 ax.set_xscale("log")
 ax.set_xlabel(r"$\gamma$")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1)
 ax.legend(loc="best",framealpha=0.5)
 ### 测试 r ，固定 gamma 为 10 ######
 rs=np.linspace(0,5)
 train_scores=[]
 test_scores=[]

for r in rs:
   regr=svm.SVR(kernel='sigmoid',coef0=r,gamma=10)
   regr.fit(X_train,y_train)
   train_scores.append(regr.score(X_train,y_train))
   test_scores.append(regr.score(X_test, y_test))
 ax=fig.add_subplot(1,2,2)
 ax.plot(rs,train_scores,label="Training score ",marker='+' )
 ax.plot(rs,test_scores,label= " Testing score ",marker='o' )
 ax.set_title( "SVR_sigmoid_r gamma=10")
 ax.set_xlabel(r"r")
 ax.set_ylabel("score")
 ax.set_ylim(-1,1)
 ax.legend(loc="best",framealpha=0.5)
 plt.show()

# 调用 test_SVR_sigmoid
test_SVR_sigmoid(X_train,X_test,y_train,y_test)

来源：https://www.cnblogs.com/tszr/p/10799437.html