基于随机森林的轴承寿命预测和故障诊断

# -*- coding: utf-8 -*-
"""
Created on Mon Mar  9 13:24:07 2020

@author: Administrator
"""
from scipy.io import loadmat
from sklearn.ensemble import RandomForestRegressor
from sklearn.neural_network import MLPRegressor
from sklearn.svm import SVR
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
from sklearn.metrics import r2_score
from sklearn import metrics

mpl.rcParams['font.sans-serif'] = ['KaiTi']
mpl.rcParams['font.serif'] = ['KaiTi']         #解决matplotlib无法显示中文的问题
plt.rcParams['axes.unicode_minus']=False       #解决负数坐标显示问题
# In[2] 原始29特征 +随机森林
data=loadmat('original_data.mat')
train_data=data['train_data']
train_label=data['train_label'].reshape(-1,1)# 数据集变一列 (-1,n)变n列
test_data=data['test_data']
test_label=data['test_label'].reshape(-1,1)
# reshape(1,-1) 数据集变一行 ,(n,-1):变n行
rf1 = RandomForestRegressor(max_depth=2, random_state=0)

rf1.fit (train_data,train_label.ravel())
test_pred1=rf1.predict(test_data).reshape(-1,1)
plt.figure()
plt.plot(test_label,'-*',label='true')
plt.plot(test_pred1,'-p',label='preict')
plt.title('29特征+随机森林')
plt.xlabel('测试样本')
plt.ylabel('剩余寿命/h')
plt.legend()
plt.show()

# 计算各种指标
print('原始29特征+随机森林')
# mape
test_mape=np.mean(np.abs((test_pred1-test_label)/test_label))
# rmse
test_rmse=np.sqrt(np.mean(np.square(test_pred1-test_label)))
# mae
test_mae=np.mean(np.abs(test_pred1-test_label))
# R2
test_r2=r2_score(test_label,test_pred1)
print('原始29特征+随机森林测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2)


# In[3] KPCA+随机森林回归

data=loadmat('kpca_data.mat')
train_data=data['train_data']
train_label=data['train_label'].reshape(-1,1)
test_data=data['test_data']
test_label=data['test_label'].reshape(-1,1)

rf2 = RandomForestRegressor(max_depth=10, random_state=0)

rf2.fit(train_data,train_label.ravel()) # ravel 多维变一维
test_pred2=rf2.predict(test_data).reshape(-1,1)
plt.figure()
plt.plot(test_label,'-*',label='true')
plt.plot(test_pred2,'-p',label='preict')
plt.title('KPCA+随机森林')
plt.xlabel('测试样本')
plt.ylabel('剩余寿命/h')
plt.legend()
plt.show()

print('kpca+随机森林')

# mape
test_mape=np.mean(np.abs((test_pred2-test_label)/test_label))
# rmse
# test_rmse=np.sqrt(np.mean(np.square(test_pred2-test_label)))
test_rmse=np.sqrt(metrics.mean_squared_error(test_label,test_pred2))
# mae
test_mae=np.mean(np.abs(test_pred2-test_label))
# R2
test_r2=r2_score(test_label,test_pred2)
print('KPCA+随机森林测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2)

# In[4] kpca+SVR回归
rf3 = SVR(C=100,gamma=1)
rf3.fit (train_data,train_label.ravel())
test_pred3=rf3.predict(test_data).reshape(-1,1)
plt.figure()
plt.plot(test_label,'-*',label='true')
plt.plot(test_pred3,'-p',label='preict')
plt.title('KPCA+SVR')
plt.xlabel('测试样本')
plt.ylabel('剩余寿命/h')
plt.legend()
plt.show()

print('kpca+SVR')
# mape
test_mape=np.mean(np.abs((test_pred3-test_label)/test_label))
# rmse
test_rmse=np.sqrt(np.mean(np.square(test_pred3-test_label)))
# mae
test_mae=np.mean(np.abs(test_pred3-test_label))
# R2
test_r2=r2_score(test_label,test_pred3)
print('KPCA+SVR测试集的mape:',test_mape,' rmse:',test_rmse,' mae:',test_mae,' R2:',test_r2)



# In[] 画在一起
plt.figure()
plt.plot(test_label,'-*',label='true')
plt.plot(test_pred1,'-p',label='随机森林')
plt.plot(test_pred2,'-s',label='KPCA+随机森林')
plt.plot(test_pred3,'-o',label='KPCA+SVR')
plt.title('各种对比')
plt.xlabel('测试样本')
plt.ylabel('剩余寿命/h')
plt.legend()
plt.show()