example 1

import pandas as pd
import numpy as np

data = pd.read_csv('refs/two_class_example.csv')

·¾¶Ò²Ö§³ÖÁªÏëÊäÈë

data.describe().pipe(print)
data.count().pipe(print)

                y          yhat
count  500.000000  5.000000e+02
mean     0.516000  5.447397e-01
std      0.500244  4.138621e-01
min      0.000000  1.794262e-07
25%      0.000000  7.289481e-02
50%      1.000000  6.569442e-01
75%      1.000000  9.794348e-01
max      1.000000  9.999965e-01
y       500
yhat    500
dtype: int64

y=1ÅжÏΪºÃÈË£¬ÏàÓ¦µØ£¬yhatÆÕ±é»á¸ß¡£

data["good"] = data.y
data["bad"] = 1 - data.y
data["score"] = data.yhat

summary[source]

summary(df, n_group=10)

Calculation KS statistic Inspired by one WenSui Liu's blog at https://statcompute.wordpress.com/2012/11/18/calculating-k-s-statistic-with-python/

Parmaters

df: pandas.DataFrame with M x N size. M length is the number of bins. N measures the number of metrics related to KS. n_group: float The number of cutted groups.

Returns

agg2 : The DataFrame return with KS and related metrics.

summary(data, n_group = 10)

	min_scr	max_scr	bads	goods	total	odds	bad_rate	ks	max_ks
0	1.794262e-07	0.002773	50	0	50	0.00	100.00%	20.66
1	2.810221e-03	0.036310	49	1	50	0.02	98.00%	40.52
2	3.670582e-02	0.122027	43	7	50	0.16	86.00%	55.58
3	1.225460e-01	0.325715	37	13	50	0.35	74.00%	65.83
4	3.269821e-01	0.655164	31	19	50	0.61	62.00%	71.27	<----
5	6.587248e-01	0.853443	22	28	50	1.27	44.00%	69.51
6	8.561391e-01	0.958957	7	43	50	6.14	14.00%	55.74
7	9.623505e-01	0.987179	1	49	50	49.00	2.00%	37.16
8	9.875471e-01	0.997897	2	48	50	24.00	4.00%	19.38
9	9.979229e-01	0.999997	0	50	50	inf	0.00%	-0.00

example 2

import pandas as pd
import numpy as np
from sklearn.metrics import roc_curve
import matplotlib.pyplot as plt

data = pd.read_csv('refs/two_class_example.csv')