2016.4.10 pca简单实践
对于一个半导体数据的处理，http://archive.ics.uci.edu/ml/machine-learning-databases/secom
其中有三个文件，一个是原始数据文件，为了测试pca进行降维使用。
一个label文件为了分类使用。
一个说明文档。
import numpy as np
from matplotlib import pyplot as plt
＃读取文件
fr = open('/Users/＊＊/Desktop/test');
delim = " ";
stringArr = [line.strip().split(delim) for line in fr.readlines()]
dataArr = [map(float,line) for line in stringArr]
numFeat = np.shape(dataArr)[1]
numSample = np.shape(dataArr)[0]
＃对于nan的缺失数据使用有效数据的平均值来代替
for i in range(numFeat):
notnan = [];
for j in range(numSample):
if ~np.isnan(dataArr[j][i]):
notnan.append(dataArr[j][i]);
meanVal = 0;
if len(notnan)!=0:
meanVal = np.mean(notnan);
for j in range(numSample):
if np.isnan(dataArr[j][i]):
dataArr[j][i] = meanVal;
＃求特征值和特征向量
meanVals = np.mean(dataArr,axis = 0)
meanRemoved = dataArr - meanVals;
covMat = np.cov(meanRemoved,rowvar = 0)
eigVals,eigVects = np.linalg.eig(covMat)
＃累计分布
s = np.sum(eigVals);
p = np.array(eigVals);
for i in range(len(eigVals)):
p[i] = np.sum(eigVals[range(i)])/s;
＃绘图
fig = plt.figure()
ax = fig.add_subplot(111)
x = range(len(p));
y = p;
ax.scatter(x,y);
plt.show();
>>> p
array([ 0. , 0.59254058, 0.83377877, 0.9252789 , 0.94828469,
0.96287661, 0.96806479, 0.97129137, 0.97443814, 0.97706893,
0.979382 , 0.98155734, 0.98363016, 0.985321 , 0.98657691,
0.98778044, 0.98892136, 0.99003252, 0.9909571 , 0.99186213,
0.99272358, 0.99346245, 0.9941864 , 0.99484841, 0.99542112,
0.99597946, 0.99648382, 0.99697404, 0.997435 , 0.99782913,
0.9981968 , 0.99849341, 0.99865661, 0.99881645, 0.99893773,
0.99905403, 0.99915929, 0.99925188, 0.99933205, 0.99939131,
0.99944629, 0.99949322, 0.99953878, 0.99957663, 0.99961259,
0.99964305, 0.99966912, 0.99969317, 0.99971385, 0.99973346,
0.99975237, 0.99977081, 0.99978788, 0.99980265, 0.99981658,
0.99982939, 0.99984186, 0.99985331, 0.99986451, 0.99987506,
0.99988515, 0.99989438, 0.99990335, 0.99991153, 0.99991916,
0.99992542, 0.99993154, 0.9999375 , 0.99994297, 0.99994756,
0.99995189, 0.99995563, 0.99995927, 0.99996269, 0.99996542,
0.99996796, 0.99997014, 0.99997209, 0.99997393, 0.99997566,
0.99997721, 0.99997864, 0.99997992, 0.99998113, 0.99998235,
0.99998347, 0.99998456, 0.99998548, 0.99998632, 0.99998708,
0.99998782, 0.9999885 , 0.99998909, 0.99998961, 0.99999011,
0.9999906 , 0.99999105, 0.99999149, 0.9999919 , 0.99999229,
0.99999265, 0.99999299, 0.99999331, 0.99999362, 0.9999939 ,
0.99999417, 0.99999443, 0.99999468, 0.99999492, 0.99999516,
0.99999538, 0.99999558, 0.99999579, 0.99999598, 0.99999616,
0.99999633, 0.99999649, 0.99999665, 0.99999681, 0.99999696,
0.99999711, 0.99999725, 0.99999738, 0.99999751, 0.99999764,
0.99999775, 0.99999786, 0.99999797, 0.99999807, 0.99999817,
0.99999827, 0.99999836, 0.99999845, 0.99999853, 0.99999861,
0.99999869, 0.99999877, 0.99999885, 0.99999892, 0.99999898,
0.99999905, 0.99999911, 0.99999916, 0.99999922, 0.99999927,
0.99999932, 0.99999936, 0.99999941, 0.99999945, 0.99999948,
0.99999952, 0.99999955, 0.99999957, 0.9999996 , 0.99999962,
0.99999964, 0.99999966, 0.99999968, 0.9999997 , 0.99999972,
0.99999974, 0.99999975, 0.99999977, 0.99999978, 0.9999998 ,
0.99999981, 0.99999982, 0.99999984, 0.99999985, 0.99999985,
0.99999986, 0.99999987, 0.99999988, 0.99999989, 0.99999989,
0.9999999 , 0.99999991, 0.99999991, 0.99999992, 0.99999992,
0.99999993, 0.99999993, 0.99999993, 0.99999994, 0.99999994,
0.99999994, 0.99999995, 0.99999995, 0.99999995, 0.99999995,
0.99999996, 0.99999996, 0.99999996, 0.99999996, 0.99999997,
0.99999997, 0.99999997, 0.99999997, 0.99999997, 0.99999997,
0.99999998, 0.99999998, 0.99999998, 0.99999998, 0.99999998,
0.99999998, 0.99999998, 0.99999998, 0.99999999, 0.99999999,
0.99999999, 0.99999999, 0.99999999, 0.99999999, 0.99999999,
0.99999999, 0.99999999, 0.99999999, 0.99999999, 0.99999999,
0.99999999, 0.99999999, 0.99999999, 0.99999999, 0.99999999,
0.99999999, 0.99999999, 0.99999999, 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ,
1. , 1. , 1. , 1. , 1. ])
参考资料：
1. 机器学习实战
2. http://archive.ics.uci.edu/ml/machine-learning-databases/secom
ps：
机器学习实践上的例子程序看明白思路后自己实现，语法不兼容。
这个程序也没有经过运行的优化，不是向量化的处理。