逻辑回归（Logistic Regression）通俗解释

-0.017612	14.053064	0
-1.395634	4.662541	1
-0.752157	6.538620	0
-1.322371	7.152853	0
0.423363	11.054677	0
0.406704	7.067335	1
0.667394	12.741452	0
-2.460150	6.866805	1
0.569411	9.548755	0
-0.026632	10.427743	0

 1 '''
 2 Created on Oct 27, 2010
 3 Logistic Regression Working Module
 4 @author: Peter
 5 '''
 6 from numpy import *
 7 import os
 8 
 9 path = 'D:\MechineLearning\MLiA_SourceCode\machinelearninginaction\Ch05'
10 training_sample = 'trainingSample.txt'
11 testing_sample = 'testingSample.txt'
12 
13 # 从文件中读入训练样本的数据，同上面给出的示例数据
14 # 下面第20行代码中的1.0表示x0 = 1
15 def loadDataSet(p, file_n):
16     dataMat = []; labelMat = []
17     fr = open(os.path.join(p, file_n))
18     for line in fr.readlines():
19         lineArr = line.strip().split()
20         dataMat.append([1.0, float(lineArr[0]), float(lineArr[1])])  # 三个特征x0, x1, x2
21         labelMat.append(int(lineArr[2]))  # 标准答案y
22     return dataMat,labelMat
23 
24 def sigmoid(inX):
25     return 1.0/(1+exp(-inX))
26 
27 # 梯度下降法求回归系数a，由于样本量少，我将迭代次数改成了1000次
28 def gradAscent(dataMatIn, classLabels):
29     dataMatrix = mat(dataMatIn)             #convert to NumPy matrix
30     labelMat = mat(classLabels).transpose() #convert to NumPy matrix
31     m,n = shape(dataMatrix)
32     alpha = 0.001  # 学习率
33     maxCycles = 1000
34     weights = ones((n,1))
35     for k in range(maxCycles):              # heavy on matrix operations
36         h = sigmoid(dataMatrix*weights)     # 模型预测值, 90 x 1
37         error = h - labelMat                # 真实值与预测值之间的误差, 90 x 1
38         temp = dataMatrix.transpose()* error # 交叉熵代价函数对所有参数的偏导数, 3 x 1
39         weights = weights - alpha * temp  # 更新权重
40     return weights
41 
42 # 下面是我自己写的测试函数
43 def test_logistic_regression():
44     dataArr, labelMat = loadDataSet(path, training_sample)  # 读入训练样本中的原始数据
45     A = gradAscent(dataArr, labelMat)  # 回归系数a的值
46     h = sigmoid(mat(dataArr)*A)  #预测结果h(a)的值
47     print(dataArr, labelMat)
48     print(A)
49     print(h)
50     # plotBestFit(A.getA())
51 
52 test_logistic_regression()

([[1.0, -0.017612, 14.053064], [1.0, -1.395634, 4.662541], [1.0, -0.752157, 6.53862], [1.0, -1.322371, 7.152853], [1.0, 0.423363, 11.054677], [1.0, 0.406704, 7.067335], [1.0, 0.667394, 12.741452], [1.0, -2.46015, 6.866805], [1.0, 0.569411, 9.548755], [1.0, -0.026632, 10.427743]], [0, 1, 0, 0, 0, 1, 0, 1, 0, 0])  

[[ 1.39174871]
[-0.5227482 ]
[-0.33100373]]  

[[ 0.03730313]
[ 0.64060602]
[ 0.40627881]
[ 0.4293251 ]
[ 0.07665396]
[ 0.23863652]
[ 0.0401329 ]
[ 0.59985228]
[ 0.11238742]
[ 0.11446212]]  

 1 # 分类效果展示，参数weights就是回归系数
 2 def plotBestFit(weights):
 3     import matplotlib.pyplot as plt
 4     dataMat,labelMat=loadDataSet()
 5     dataArr = array(dataMat)
 6     n = shape(dataArr)[0]
 7     xcord1 = []; ycord1 = []
 8     xcord2 = []; ycord2 = []
 9     for i in range(n):
10         if int(labelMat[i])== 1:
11             xcord1.append(dataArr[i,1]); ycord1.append(dataArr[i,2])
12         else:
13             xcord2.append(dataArr[i,1]); ycord2.append(dataArr[i,2])
14     fig = plt.figure()
15     ax = fig.add_subplot(111)
16     ax.scatter(xcord1, ycord1, s=30, c='red', marker='s')
17     ax.scatter(xcord2, ycord2, s=30, c='green')
18     x = arange(-3.0, 3.0, 0.1)
19     y = (-weights[0]-weights[1]*x)/weights[2]  # x2 = f(x1)
20     ax.plot(x, y)
21     plt.xlabel('X1'); plt.ylabel('X2');
22     plt.show()

[[ 5.262118 ]
[ 0.60847797]
[-0.75168429]]

1 def predict_test_sample():
2     A = [5.262118, 0.60847797, -0.75168429]  # 上面计算出来的回归系数a
3     dataArr, labelMat = loadDataSet(path, testing_sample)  
4     h_test = sigmoid(mat(dataArr) * mat(A).transpose())  # 将读入的数据和A转化成numpy中的矩阵
5     print(h_test)  # 预测的结果

[[ 0.99714035]
[ 0.04035907]
[ 0.12535895]
[ 0.99048731]
[ 0.98075409]
[ 0.97708653]
[ 0.09004989]
[ 0.97884487]
[ 0.28594188]
[ 0.00359693]]

0.089392	-0.715300	1
1.825662	12.693808	0
0.197445	9.744638	0
0.126117	0.922311	1
-0.679797	1.220530	1
0.677983	2.556666	1
0.761349	10.693862	0
-2.168791	0.143632	1
1.388610	9.341997	0
0.317029	14.739025	0