divide by zero encountered in log2 に関連する問題を解決したい

前提・実現したいこと

RuntimeWarning: divide by zero encountered in log2
RuntimeWarning: invalid value encountered in double_scalars
を解決したい

発生している問題・エラーメッセージ

sample-kadai5.py:54: RuntimeWarning: divide by zero encountered in log2
  gain = En - (-1*((Beta0/8)*((Xw0/Beta0)*np.log2(Xw0/Beta0)+(Yw0/Beta0)*np.log2(Yw0/Beta0))+(Beta1/8)*((Xw1/Beta1)*np.log2(Xw1/Beta1)+(Yw1/Beta1)*np.log2(Yw1/Beta1))+(Beta2/8)*((Xw2/Beta2)*np.log2(Xw2/Beta2)+(Yw2/Beta2)*np.log2(Yw2/Beta2))))
sample-kadai5.py:54: RuntimeWarning: invalid value encountered in double_scalars
  gain = En - (-1*((Beta0/8)*((Xw0/Beta0)*np.log2(Xw0/Beta0)+(Yw0/Beta0)*np.log2(Yw0/Beta0))+(Beta1/8)*((Xw1/Beta1)*np.log2(Xw1/Beta1)+(Yw1/Beta1)*np.log2(Yw1/Beta1))+(Beta2/8)*((Xw2/Beta2)*np.log2(Xw2/Beta2)+(Yw2/Beta2)*np.log2(Yw2/Beta2))))

該当のソースコード

Python3
1import numpy as np
2import pandas as pd
3
4#圧力の低を０高を１に、煙の色の白を０黒を１赤を２に、鋼の色の青を０茶を１に、状態の正常を０異常１に対応
5plantdata = pd.DataFrame(data={'pressure':[0,1,1,1,0,1,0,1],'smokec':[0,1,1,2,1,0,0,0],'steelc':[0,1,0,0,0,1,1,0],'state':[0,1,1,0,1,1,1,0]},columns=['pressure','smokec','steelc','state'])
6rg = np.zeros(3)
7
8#平均情報量を計算するためのまとまり
9x_bool = plantdata['state'] == 1
10x = x_bool.sum()
11y = 8 - x
12En = -(x/8)*np.log2(x/8)-(y/8)*np.log2(y/8)
13
14#コラムに関する情報利得と情報利得費を計算するためのまとまり
15tc = 0
16for column_name in plantdata:
17 epsilon = (10)**(-1)
18 #コラム0,1,2が何個あるか計算
19 beta_0_bool = plantdata[column_name] == 0
20 Beta0 = beta_0_bool.sum() + epsilon
21 beta_1_bool = plantdata[column_name] == 1
22 Beta1 = beta_1_bool.sum() + epsilon
23 beta_2_bool = plantdata[column_name] == 2
24 Beta2 = beta_2_bool.sum() + epsilon
25
26 #コラムをそれぞれ0,1,2に対して抽出
27 beta_0 = plantdata[plantdata[column_name] == 0]
28 beta_1 = plantdata[plantdata[column_name] == 1]
29 beta_2 = plantdata[plantdata[column_name] == 2]
30
31 #indexがばらばらになっているので揃える
32 BW0 = beta_0.reset_index(drop=True)
33 BW1 = beta_1.reset_index(drop=True)
34 BW2 = beta_2.reset_index(drop=True)
35
36 #コラムが0のものについて計算に必要な値を出す
37 Xw0_bool = BW0['state'] == 1
38 Xw0 = Xw0_bool.sum()
39 Yw0 = Beta0 - Xw0
40
41 #同様に1
42 Xw1_bool = BW1['state'] == 1
43 Xw1 = Xw1_bool.sum()
44 Yw1 = Beta1 - Xw1
45
46 #同様に2
47 Xw2_bool = BW2['state'] == 1
48 Xw2 = Xw2_bool.sum()
49 Yw2 = Beta2 - Xw2
50
51 #情報利得の計算
52 gain = En - (-1*((Beta0/8)*((Xw0/Beta0)*np.log2(Xw0/Beta0)+(Yw0/Beta0)*np.log2(Yw0/Beta0))+(Beta1/8)*((Xw1/Beta1)*np.log2(Xw1/Beta1)+(Yw1/Beta1)*np.log2(Yw1/Beta1))+(Beta2/8)*((Xw2/Beta2)*np.log2(Xw2/Beta2)+(Yw2/Beta2)*np.log2(Yw2/Beta2))))
53 rg[tc] = gain
54 tc = tc + 1
55
56#出力
57print(En)
58print(rg)
59