Python实现PCA

糖逗

参考书籍：《机器学习实战》

1. import numpy as np
   
2. 
   
3. def loadDataSet(fileName, delim = '\t'):
   
4.     fr = open(fileName)
   
5.     stringArr = [line.strip().split(delim) for line in fr.readlines()]
   
6.     datArr = [list(map(float,line)) for line in stringArr]
   
7.     return np.mat(datArr)
   
8. 
   
9. def pca(dataMat, topNfeat = 9999999):
   
10.     meanVals = np.mean(dataMat, axis = 0)
    
11.     meanRemoved = dataMat - meanVals
    
12.     covMat = np.cov(meanRemoved, rowvar = 0)#rowvar=0表示传入的一行表示一个样本
    
13.     eigVals, eigVects = np.linalg.eig(np.mat(covMat))
    
14.     eigValInd = np.argsort(eigVals)           
    
15.     eigValInd = eigValInd[:-(topNfeat+1):-1]  
    
16.     redEigVects = eigVects[:,eigValInd]      
    
17.     lowDDataMat = meanRemoved * redEigVects#所有点在新的一组基上的值，无方向
    
18.     reconMat = (lowDDataMat * redEigVects.T) + meanVals#所有点在新的一组基上的投影，有方向的
    
19.     return lowDDataMat, reconMat
    
20. 
    
21. 
    
22. if __name__ == '__main__':
    
23.     dataMat = loadDataSet(r'C:\...\testSet.txt')
    
24.     lowData, reconMat = pca(dataMat, 1)
    
25.    
    
26.     import matplotlib.pyplot as plt
    
27.     fig = plt.figure()
    
28.     ax = fig.add_subplot(111)
    
29.     ax.scatter(dataMat[:, 0].flatten().A[0], dataMat[:, 1].flatten().A[0],
    
30.                marker = '^', s = 90)
    
31.     ax.scatter(reconMat[:, 0].flatten().A[0], reconMat[:, 1].flatten().A[0],
    
32.                marker = 'o', s = 50, c = 'red')
    
33.    

复制代码

糖逗