:param count: DataFrame 待分箱变量各取值的正负样本数
:param t_num_0:
:param t_num_1:
:return: 该分箱的ks值
计算公式：KS_i = |sum_i / sum_T - (size_i - sum_i)/ (size_T - sum_T)|
"""
# 计算左评分区间的累计好账户数占总好账户数比率（good %)和累计坏账户数占总坏账户数比率（bad %）。
good_left = count[1].sum() / t_num_1 if count[1].sum() != 0 else 1
bad_left = count[0].sum() / t_num_0 if count[0].sum() != 0 else 1
return abs(good_left - bad_left)
if __name__ == '__main__':
e_data = ExampleData()
df = e_data.get_iris2()
print(df.head())
me_bin = BestDistinguishBinning('y', 'ks')
me_bin.max_bin = 4
# count = pd.crosstab(df['sepal_width'], df['y'])
# print(MinEntropyBinning.calc_entropy(count))
print('切分点集合', me_bin.calc(df, 'sepal_width'))
# @Author : HD
import sys
from io import StringIO
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import SGDClassifier
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import GradientBoostingRegressor
from utils.example_data import ExampleData
if __name__ == '__main__':
"""
截取，模型训练的中间过程。
"""
edd = ExampleData()
df = edd.get_barest_cancer_data()
cols = df.columns
old_stdout = sys.stdout
sys.stdout = mystdout = StringIO()
clf = GradientBoostingRegressor(verbose=1)
clf.fit(df[cols[:-2]], df[cols[-1]])
sys.stdout = old_stdout
loss_history = mystdout.getvalue()
print(type(loss_history))
# for i in loss_history.split('\n'):
# print(i, len(i))
print([{
'epoch': j[7:10].strip(' '),