机器学习--Logistic回归

 1 # -*- coding:utf-8 -*-
 2 import numpy as np
 3 
 4 def load_data(file_name):
 5     # 载入数据
 6     data_mat = []
 7     labels = []
 8 
 9     with open(file_name) as file:
10         for line in file.readlines():
11             line_arr = line.strip().split(‘,‘)
12             data_mat.append([float(line_arr[0]), float(line_arr[1]), float(line_arr[2])])
13             labels.append(int(line_arr[3]))
14     return data_mat, labels
15 
16 
17 def sigmoid(x):
18     # 阶跃函数
19     if -x > np.log(np.finfo(type(x)).max):
20         return 0.0
21     else:
22         return 1.0 / (1 + np.exp(-x))
23 
24 
25 def grad_ascent(data_mat, data_labels, num_iter=200):
26     """随机梯度上升算法"""
27     data_mat = np.array(data_mat)
28     m, n = np.shape(data_mat)
29 
30     weights = np.ones(n).astype(np.float)
31     for j in range(num_iter):
32         data_index = list(range(m))
33         for i in range(m):
34             alpha = 0.001 + 4 / (1.0 + j + i)
35 
36             random_index = int(np.random.uniform(0, len(data_index)))
37             h = sigmoid(sum(data_mat[random_index] * weights))
38             error = data_labels[random_index] - h
39             weights = weights + alpha * error * data_mat[random_index]
40             del (data_index[random_index])
41 
42     return weights
43 
44 
45 def test(x, name):
46     file_name = name
47     data_mat, labels = load_data(file_name)
48     weights = grad_ascent(data_mat, labels)
49     print(weights)
50     res = classify_vector(x, weights)
51     print(res)
52 
53 
54 def classify_vector(inx, weights):
55     prob = sigmoid(sum(inx * weights))
56     if prob > 0.5:
57         return 1.0
58     else:
59         return 0.0
60 
61 
62 name = ‘data/haberman.txt‘
63 test([72, 63, 0], name)