forked from joymallyac/FairSSL
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathMeasure.py
More file actions
180 lines (145 loc) · 8.05 KB
/
Copy pathMeasure.py
File metadata and controls
180 lines (145 loc) · 8.05 KB
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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
import numpy as np
import copy,math
from sklearn.metrics import confusion_matrix,classification_report
from sklearn.neighbors import KDTree
from sklearn.neighbors import NearestNeighbors
def get_counts(clf, x_train, y_train, x_test, y_test, test_df, biased_col, metric='aod'):
# clf.fit(x_train, y_train)
y_pred = clf.predict(x_test)
cnf_matrix = confusion_matrix(y_test, y_pred)
TN, FP, FN, TP = confusion_matrix(y_test,y_pred).ravel()
test_df_copy = copy.deepcopy(test_df)
test_df_copy['current_pred_' + biased_col] = y_pred
test_df_copy['TP_' + biased_col + "_1"] = np.where((test_df_copy['Probability'] == 1) &
(test_df_copy['current_pred_' + biased_col] == 1) &
(test_df_copy[biased_col] == 1), 1, 0)
test_df_copy['TN_' + biased_col + "_1"] = np.where((test_df_copy['Probability'] == 0) &
(test_df_copy['current_pred_' + biased_col] == 0) &
(test_df_copy[biased_col] == 1), 1, 0)
test_df_copy['FN_' + biased_col + "_1"] = np.where((test_df_copy['Probability'] == 1) &
(test_df_copy['current_pred_' + biased_col] == 0) &
(test_df_copy[biased_col] == 1), 1, 0)
test_df_copy['FP_' + biased_col + "_1"] = np.where((test_df_copy['Probability'] == 0) &
(test_df_copy['current_pred_' + biased_col] == 1) &
(test_df_copy[biased_col] == 1), 1, 0)
test_df_copy['TP_' + biased_col + "_0"] = np.where((test_df_copy['Probability'] == 1) &
(test_df_copy['current_pred_' + biased_col] == 1) &
(test_df_copy[biased_col] == 0), 1, 0)
test_df_copy['TN_' + biased_col + "_0"] = np.where((test_df_copy['Probability'] == 0) &
(test_df_copy['current_pred_' + biased_col] == 0) &
(test_df_copy[biased_col] == 0), 1, 0)
test_df_copy['FN_' + biased_col + "_0"] = np.where((test_df_copy['Probability'] == 1) &
(test_df_copy['current_pred_' + biased_col] == 0) &
(test_df_copy[biased_col] == 0), 1, 0)
test_df_copy['FP_' + biased_col + "_0"] = np.where((test_df_copy['Probability'] == 0) &
(test_df_copy['current_pred_' + biased_col] == 1) &
(test_df_copy[biased_col] == 0), 1, 0)
a = test_df_copy['TP_' + biased_col + "_1"].sum()
b = test_df_copy['TN_' + biased_col + "_1"].sum()
c = test_df_copy['FN_' + biased_col + "_1"].sum()
d = test_df_copy['FP_' + biased_col + "_1"].sum()
e = test_df_copy['TP_' + biased_col + "_0"].sum()
f = test_df_copy['TN_' + biased_col + "_0"].sum()
g = test_df_copy['FN_' + biased_col + "_0"].sum()
h = test_df_copy['FP_' + biased_col + "_0"].sum()
if metric=='aod':
return calculate_average_odds_difference(a, b, c, d, e, f, g, h)
elif metric=='eod':
return calculate_equal_opportunity_difference(a, b, c, d, e, f, g, h)
elif metric=='recall':
return calculate_recall(TP,FP,FN,TN)
elif metric=='far':
return calculate_far(TP,FP,FN,TN)
elif metric=='precision':
return calculate_precision(TP,FP,FN,TN)
elif metric=='accuracy':
return calculate_accuracy(TP,FP,FN,TN)
elif metric=='F1':
return calculate_F1(TP,FP,FN,TN)
elif metric=='TPR':
return calculate_TPR_difference(a, b, c, d, e, f, g, h)
elif metric=='FPR':
return calculate_FPR_difference(a, b, c, d, e, f, g, h)
elif metric == "DI":
return calculate_Disparate_Impact(a, b, c, d, e, f, g, h)
elif metric == "SPD":
return calculate_SPD(a, b, c, d, e, f, g, h)
def calculate_average_odds_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
# TPR_male = TP_male/(TP_male+FN_male)
# TPR_female = TP_female/(TP_female+FN_female)
# FPR_male = FP_male/(FP_male+TN_male)
# FPR_female = FP_female/(FP_female+TN_female)
# average_odds_difference = abs(abs(TPR_male - TPR_female) + abs(FPR_male - FPR_female))/2
FPR_diff = calculate_FPR_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female)
TPR_diff = calculate_TPR_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female)
average_odds_difference = (FPR_diff + TPR_diff)/2
#print("average_odds_difference",average_odds_difference)
return round(average_odds_difference,2)
def calculate_Disparate_Impact(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
P_male = (TP_male + FP_male)/(TP_male + TN_male + FN_male + FP_male)
P_female = (TP_female + FP_female)/(TP_female + TN_female + FN_female + FP_female)
DI = (P_female/P_male)
return round((1 - abs(DI)),2)
def calculate_SPD(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
P_male = (TP_male + FP_male)/(TP_male + TN_male + FN_male + FP_male)
P_female = (TP_female + FP_female) /(TP_female + TN_female + FN_female + FP_female)
SPD = (P_female - P_male)
return round(abs(SPD),2)
def calculate_equal_opportunity_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
# TPR_male = TP_male/(TP_male+FN_male)
# TPR_female = TP_female/(TP_female+FN_female)
# equal_opportunity_difference = abs(TPR_male - TPR_female)
#print("equal_opportunity_difference:",equal_opportunity_difference)
return calculate_TPR_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female)
def calculate_TPR_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
TPR_male = TP_male/(TP_male+FN_male)
TPR_female = TP_female/(TP_female+FN_female)
# print("TPR_male:",TPR_male,"TPR_female:",TPR_female)
diff = (TPR_male - TPR_female)
return round(diff,2)
def calculate_FPR_difference(TP_male , TN_male, FN_male,FP_male, TP_female , TN_female , FN_female, FP_female):
FPR_male = FP_male/(FP_male+TN_male)
FPR_female = FP_female/(FP_female+TN_female)
# print("FPR_male:",FPR_male,"FPR_female:",FPR_female)
diff = (FPR_female - FPR_male)
return round(diff,2)
def calculate_recall(TP,FP,FN,TN):
if (TP + FN) is not 0:
recall = TP / (TP + FN)
else:
recall = 0
return round(recall,2)
def calculate_far(TP,FP,FN,TN):
if (FP + TN) is not 0:
far = FP / (FP + TN)
else:
far = 0
return round(far,2)
def calculate_precision(TP,FP,FN,TN):
if (TP + FP) is not 0:
prec = TP / (TP + FP)
else:
prec = 0
return round(prec,2)
def calculate_F1(TP,FP,FN,TN):
precision = calculate_precision(TP,FP,FN,TN)
recall = calculate_recall(TP,FP,FN,TN)
F1 = (2 * precision * recall)/(precision + recall)
return round(F1,2)
def calculate_accuracy(TP,FP,FN,TN):
return round((TP + TN)/(TP + TN + FP + FN),2)
def consistency_score(X, y, n_neighbors=5):
num_samples = X.shape[0]
# y = y.values # Do it if it's not np array
# learn a KNN on the features
nbrs = NearestNeighbors(n_neighbors, algorithm='ball_tree').fit(X)
_, indices = nbrs.kneighbors(X)
# compute consistency score
consistency = 0.0
for i in range(num_samples):
consistency += np.abs(y[i] - np.mean(y[indices[i]]))
consistency = 1.0 - consistency/num_samples
return consistency
def measure_final_score(test_df, clf, X_train, y_train, X_test, y_test, biased_col, metric):
df = copy.deepcopy(test_df)
return get_counts(clf, X_train, y_train, X_test, y_test, df, biased_col, metric=metric)