1 #!/usr/bin/env python
 2 # encoding: utf-8
 3 
 4 """
 5 @version: python37
 6 @author: Geoffrey
 7 @file: KNN.py
 8 @time: 18-11-12 下午2:53
 9 """
10 
11 ‘‘‘
12 需求：使用KNN预测分类
13 样本集：
14 1,  1,  a
15 1,  2,  a
16 1.5,1.5,a
17 3,  4,  b
18 4,  4,  b
19 预测数据：
20 2, 2
21 ‘‘‘
22 
23 import numpy as np
24 import pandas as pd
25 import matplotlib.pyplot as plt
26 # import matplotlib.font_manager as fm
27 
28 # 计算距离函数
29 def Minkowski_distication(data, point, m=2):
30     num, dim = data.shape
31     result = 0
32     if len(point) == dim:
33         for i in range(dim):
34             result += ((abs(data[i] - point[i])) ** m)
35         return result ** (1 / m)
36 
37 # KNN分类
38 def KNN_classify(data, distance, k):
39     index = np.argsort(distance, axis=0)
40     data[‘distance‘]  =  index
41     data = data.sort_values(‘distance‘).head(k)[2]
42 
43     return data.value_counts().head(1).index
44 
45 if __name__ == ‘__main__‘:
46     sample = [
47         [1, 1, ‘a‘],
48         [1, 2, ‘a‘],
49         [1.5, 1.5, ‘a‘],
50         [3, 4, ‘b‘],
51         [4, 4, ‘b‘],
52     ]
53     df_sample = pd.DataFrame(sample)
54     point = (2, 2)
55 
56     distance = Minkowski_distication(df_sample.drop(2, axis=1), point=point)
57     result = KNN_classify(df_sample, distance, k=4)
58 
59     # 绘图
60     fig = plt.figure()
61     plt.scatter(df_sample[0], df_sample[1])
62     plt.scatter(*point, color=‘red‘)
63     plt.show()