decisionTree/GainRatio.py at master · fanny-riols/decisionTree · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
import math
from Algo import *

def Entropy(counts):
    # Pour un type en particulier
    entropy = 0

    for i in range(len(counts)):
        num = 0.0
        for j in range(len(counts[i])):
            num += counts[i][j]
        for j in range(len(counts[i])):
            if num != 0 and counts[i][j] != 0:
                a = counts[i][j] / num
                entropy -= float(a * math.log(a, 2))

    return entropy

def Gini(counts):
    # Pour un type en particulier
    gini = 1

    for i in range(len(counts)):
        num = 0.0
        for j in range(len(counts[i])):
            num += counts[i][j]
        for j in range(len(counts[i])):
            if num != 0 and counts[i][j] != 0:
                a = counts[i][j] / num
                gini -= float(a * a)

    return gini


def Information(counts, algo):
    # Pour un type en particulier
    info = 0

    total = 0
    for i in range(len(counts)):
        for j in range(len(counts[i])):
            total += counts[i][j]


    for i in range(len(counts)):
        num = 0.0
        impurity = []

        for j in range(len(counts[i])):
            num += counts[i][j]
        c = []
        for j in range(len(counts[i])):
            if num != 0:
                c.append(counts[i][j] / num)
            else:
                c.append(0)

        if algo.criteria == Criteria.Type['Entropy']:
            impurity = Entropy([c])
        elif algo.criteria == Criteria.Type['Gini']:
            impurity = Gini([c])

        a = num / total
        info += a * impurity


    return info

def InfoGain(counts, numSamples, algo):
    # Pour un type en particulier

    S = []
    total = 0
    numUnknown = 0
    for i in range(len(counts[0])):
        S.append(0)
    for i in range(len(counts)):
        for j in range(len(counts[i])):
            S[j] += counts[i][j]
            total += counts[i][j]
    for i in range(len(S)):
        S[i] = float(S[i]) / total

    impurity = []

    if algo.criteria == Criteria.Type['Entropy']:
        impurity = Entropy([S])
    elif algo.criteria == Criteria.Type['Gini']:
        impurity = Gini([S])

    print('Impurity: ', impurity)

    info = Information(counts, algo)

    print('Info: ', info)

    gain = (float(total) / numSamples) * (impurity - info)

    return gain


def SplitInformation(counts, algo):
    # Pour un type en particulier

    splitInfo = 0

    num = []
    total = 0
    for i in range(len(counts)):
        num.append(0)
        for j in range(len(counts[i])):
            num[i] += counts[i][j]
            total += counts[i][j]

    for i in range(len(num)):
        num[i] = float(num[i]) / total


    if algo.criteria == Criteria.Type['Entropy']:
        splitInfo = Entropy([num])
    elif algo.criteria == Criteria.Type['Gini']:
        splitInfo = Gini([num])

    return splitInfo


def GainRatio(counts, numSamples, algo):
    # Pour un type en particulier

    gain = InfoGain(counts, numSamples, algo)
    splInfo = SplitInformation(counts, algo)

    print(counts)
    print('gain: ', gain)
    print('splInfo: ', splInfo)

    if splInfo != 0:
        gainRatio = gain / splInfo
    else:
        gainRatio = 0


    return gainRatio