Python/machine_learning/scoring_functions.py at master · cureprotocols/Python · GitHub

Name: Python/machine_learning/scoring_functions.py at master · cureprotocols/Python · GitHub
Rating: 4.7 (4516 reviews)
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
importnumpyasnp

""" Here I implemented the scoring functions.
 MAE, MSE, RMSE, RMSLE are included.

 Those are used for calculating differences between
 predicted values and actual values.

 Metrics are slightly differentiated. Sometimes squared, rooted,
 even log is used.

 Using log and roots can be perceived as tools for penalizing big
 errors. However, using appropriate metrics depends on the situations,
 and types of data
"""


# Mean Absolute Error
defmae(predict, actual):
"""
 Examples(rounded for precision):
 >>> actual = [1,2,3];predict = [1,4,3]
 >>> float(np.around(mae(predict,actual),decimals = 2))
 0.67

 >>> actual = [1,1,1];predict = [1,1,1]
 >>> float(mae(predict,actual))
 0.0
 """
predict=np.array(predict)
actual=np.array(actual)

difference=abs(predict-actual)
score=difference.mean()

returnscore


# Mean Squared Error
defmse(predict, actual):
"""
 Examples(rounded for precision):
 >>> actual = [1,2,3];predict = [1,4,3]
 >>> float(np.around(mse(predict,actual),decimals = 2))
 1.33

 >>> actual = [1,1,1];predict = [1,1,1]
 >>> float(mse(predict,actual))
 0.0
 """
predict=np.array(predict)
actual=np.array(actual)

difference=predict-actual
square_diff=np.square(difference)

score=square_diff.mean()
returnscore


# Root Mean Squared Error
defrmse(predict, actual):
"""
 Examples(rounded for precision):
 >>> actual = [1,2,3];predict = [1,4,3]
 >>> float(np.around(rmse(predict,actual),decimals = 2))
 1.15

 >>> actual = [1,1,1];predict = [1,1,1]
 >>> float(rmse(predict,actual))
 0.0
 """
predict=np.array(predict)
actual=np.array(actual)

difference=predict-actual
square_diff=np.square(difference)
mean_square_diff=square_diff.mean()
score=np.sqrt(mean_square_diff)
returnscore


# Root Mean Square Logarithmic Error
defrmsle(predict, actual):
"""
 Examples(rounded for precision):
 >>> float(np.around(rmsle(predict=[10, 2, 30], actual=[10, 10, 30]), decimals=2))
 0.75

 >>> float(rmsle(predict=[1, 1, 1], actual=[1, 1, 1]))
 0.0
 """
predict=np.array(predict)
actual=np.array(actual)

log_predict=np.log(predict+1)
log_actual=np.log(actual+1)

difference=log_predict-log_actual
square_diff=np.square(difference)
mean_square_diff=square_diff.mean()

score=np.sqrt(mean_square_diff)

returnscore


# Mean Bias Deviation
defmbd(predict, actual):
"""
 This value is Negative, if the model underpredicts,
 positive, if it overpredicts.

 Example(rounded for precision):

 Here the model overpredicts
 >>> actual = [1,2,3];predict = [2,3,4]
 >>> float(np.around(mbd(predict,actual),decimals = 2))
 50.0

 Here the model underpredicts
 >>> actual = [1,2,3];predict = [0,1,1]
 >>> float(np.around(mbd(predict,actual),decimals = 2))
 -66.67
 """
predict=np.array(predict)
actual=np.array(actual)

difference=predict-actual
numerator=np.sum(difference) /len(predict)
denumerator=np.sum(actual) /len(predict)
# print(numerator, denumerator)
score=float(numerator) /denumerator*100

returnscore


defmanual_accuracy(predict, actual):
returnnp.mean(np.array(actual) ==np.array(predict))