回归支持向量机的PYTHON代码，支持向量机svm

# -*- coding: utf-8 -*- """ Created on Fri Apr 20 10:20:51 2018 @author: kly qq392503054 """ # 支持向量机 import pandas as pd import numpy as np import matplotlib.pyplot as plt from sklearn import svm from sklearn.preprocessing import MinMaxScaler,StandardScaler from sklearn.metrics import r2_score from scipy.io import loadmat # In[] # # 导入数据 data=loadmat('maydataa.mat')['data'] #1-10列为输入 11列为输出 in_=data[:,:-1] out_=data[:,-1].reshape(-1,1) # 归一化 ss_X=MinMaxScaler(feature_range=(-1,1)).fit(in_) ss_Y=MinMaxScaler(feature_range=(-1,1)).fit(out_) input_data = ss_X.fit_transform(in_) output_data = ss_Y.fit_transform(out_) # 前1/3组训练 后2/3组测试 n_samples=input_data.shape[0] n=np.arange(n_samples) m=int(n_samples/3) xtrain=input_data[n[0:m],:] ytrain=output_data[n[0:m],0] xtest=input_data[n[m:],:] ytest=output_data[n[m:],0] # svm 回归 clf = svm.SVR(C=1, gamma=1,kernel='rbf') clf.fit(xtrain,ytrain) result0 = clf.predict(xtrain).reshape(-1,1) result = clf.predict(xtest).reshape(-1,1) # 反归一化 train_label=ss_Y.inverse_transform(ytrain.reshape(-1,1)) test_label=ss_Y.inverse_transform(ytest.reshape(-1,1)) train_pred=ss_Y.inverse_transform(result0) test_pred=ss_Y.inverse_transform(result) # 结果统计 plt.figure() plt.subplot(1,2,1) plt.title("train set") plt.xlabel("train_sample", size=14) plt.ylabel("value", size=14) plt.plot(train_pred, color='g',label='pred',linewidth=3) plt.plot(train_label, color='r',label='true',linewidth=3) plt.legend() plt.show() plt.subplot(1,2,2) plt.title("test set") plt.xlabel("test_sample", size=14) plt.ylabel("value", size=14) plt.plot(test_pred, color='g',label='pred',linewidth=3) plt.plot(test_label, color='r',label='true',linewidth=3) plt.legend() plt.show() # mape train_mape=np.mean(np.abs((train_pred-train_label)/train_label)) # rmse train_rmse=np.sqrt(np.mean(np.square(train_pred-train_label))) # mae train_mae=np.mean(np.abs(train_pred-train_label)) # R2 train_r2=r2_score(train_label,train_pred) print('SVM训练集的mape:',train_mape,' rmse:',train_rmse,' mae:',train_mae,' R2:',train_r2) # mape test_mape=np.mean(np.abs((test_pred-test_label)/test_label)) # rmse test_rmse=np.sqrt(np.mean(np.square(test_pred-test_label))) # mae test_mae=np.mean(np.abs(test_pred-test_label)) # R2 test_r2=r2_score(test_label,test_pred) print('SVM测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2) # In[] colum11=np.vstack([train_pred,test_pred])# 用预测得到的数据 替换原来的11列进行建模 ss_X=MinMaxScaler(feature_range=(-1,1)).fit(in_) ss_Y2=MinMaxScaler(feature_range=(-1,1)).fit(colum11)# 进行归一化 input_data = ss_X.fit_transform(in_) output_data11 = ss_Y.fit_transform(colum11) # 前1/3组训练 后2/3组测试 n_samples=input_data.shape[0] n=np.arange(n_samples) m=int(n_samples/3) xtrain=input_data[n[0:m],:] ytrain=output_data11[n[0:m],0] xtest=input_data[n[m:],:] ytest=output_data11[n[m:],0] # svm 回归 clf = svm.SVR(C=1, gamma=1,kernel='rbf') clf.fit(xtrain,ytrain) result1 = clf.predict(xtrain).reshape(-1,1) result2 = clf.predict(xtest).reshape(-1,1) # 反归一化 train_label=ss_Y2.inverse_transform(ytrain.reshape(-1,1)) test_label=ss_Y2.inverse_transform(ytest.reshape(-1,1)) train_pred=ss_Y2.inverse_transform(result1) test_pred=ss_Y2.inverse_transform(result2) # 结果统计 plt.figure() plt.subplot(1,2,1) plt.title("train set") plt.xlabel("train_sample", size=14) plt.ylabel("value", size=14) plt.plot(train_pred, color='g',label='pred',linewidth=3) plt.plot(train_label, color='r',label='true',linewidth=3) plt.legend() plt.show() plt.subplot(1,2,2) plt.title("test set") plt.xlabel("test_sample", size=14) plt.ylabel("value", size=14) plt.plot(test_pred, color='g',label='pred',linewidth=3) plt.plot(test_label, color='r',label='true',linewidth=3) plt.legend() plt.show() # mape train_mape=np.mean(np.abs((train_pred-train_label)/train_label)) # rmse train_rmse=np.sqrt(np.mean(np.square(train_pred-train_label))) # mae train_mae=np.mean(np.abs(train_pred-train_label)) # R2 train_r2=r2_score(train_label,train_pred) print('SVM训练集的mape:',train_mape,' rmse:',train_rmse,' mae:',train_mae,' R2:',train_r2) # mape test_mape=np.mean(np.abs((test_pred-test_label)/test_label)) # rmse test_rmse=np.sqrt(np.mean(np.square(test_pred-test_label))) # mae test_mae=np.mean(np.abs(test_pred-test_label)) # R2 test_r2=r2_score(test_label,test_pred) print('SVM测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2) # In[] in_2=np.hstack([in_,colum11])# 用预测得到的数据 加上之前的10列输出构成11列输出 #输出依旧为原来的第11列 ss_X=MinMaxScaler(feature_range=(-1,1)).fit(in_2) ss_Y3=MinMaxScaler(feature_range=(-1,1)).fit(out_)# 进行归一化 input_data = ss_X.fit_transform(in_) output_data11 = ss_Y.fit_transform(out_) # 前1/3组训练 后2/3组测试 n_samples=input_data.shape[0] n=np.arange(n_samples) m=int(n_samples/3) xtrain=input_data[n[0:m],:] ytrain=output_data11[n[0:m],0] xtest=input_data[n[m:],:] ytest=output_data11[n[m:],0] # svm 回归 clf = svm.SVR(C=1, gamma=0.1,kernel='rbf') clf.fit(xtrain,ytrain) result1 = clf.predict(xtrain).reshape(-1,1) result2 = clf.predict(xtest).reshape(-1,1) # 反归一化 train_label=ss_Y3.inverse_transform(ytrain.reshape(-1,1)) test_label=ss_Y3.inverse_transform(ytest.reshape(-1,1)) train_pred=ss_Y3.inverse_transform(result1) test_pred=ss_Y3.inverse_transform(result2) # 结果统计 plt.figure() plt.subplot(1,2,1) plt.title("train set") plt.xlabel("train_sample", size=14) plt.ylabel("value", size=14) plt.plot(train_pred, color='g',label='pred',linewidth=3) plt.plot(train_label, color='r',label='true',linewidth=3) plt.legend() plt.show() plt.subplot(1,2,2) plt.title("test set") plt.xlabel("test_sample", size=14) plt.ylabel("value", size=14) plt.plot(test_pred, color='g',label='pred',linewidth=3) plt.plot(test_label, color='r',label='true',linewidth=3) plt.legend() plt.show() # mape train_mape=np.mean(np.abs((train_pred-train_label)/train_label)) # rmse train_rmse=np.sqrt(np.mean(np.square(train_pred-train_label))) # mae train_mae=np.mean(np.abs(train_pred-train_label)) # R2 train_r2=r2_score(train_label,train_pred) print('SVM训练集的mape:',train_mape,' rmse:',train_rmse,' mae:',train_mae,' R2:',train_r2) # mape test_mape=np.mean(np.abs((test_pred-test_label)/test_label)) # rmse test_rmse=np.sqrt(np.mean(np.square(test_pred-test_label))) # mae test_mae=np.mean(np.abs(test_pred-test_label)) # R2 test_r2=r2_score(test_label,test_pred) print('SVM测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2)