该博客介绍了如何利用LightGBM建立信用评分卡模型，通过去除共线性、变量选择来提升模型性能。文章对比了LightGBM与逻辑回归（LR）的效果，并展示了LightGBM在训练集和验证集上的KS和AUC指标。此外，还探讨了评分卡公式的变形，并给出了模型预测分数的计算方法，最终生成了评估报告。
xgb依然要去除共线性、变量选择
lr bivar要严格单调，xgb、lightGBM不需要
LightGBM评分卡
import pandas as pd
from sklearn.metrics import roc_auc_score,roc_curve,auc
from sklearn.model_selection import train_test_split
from sklearn import metrics
from sklearn.linear_model import LogisticRegression
import numpy as np
import random
import math
import time
import lightgbm as lgb
data = pd.read_csv('Bcard.txt')
data.head()
data.shape
#(95806, 13)
#看一下月份分布，我们用最后一个月做为跨时间验证集合
data.obs_mth.unique()
#array(['2018-10-31', '2018-07-31', '2018-09-30', '2018-06-30',
# '2018-11-30'], dtype=object)
df_train = data[data.obs_mth != '2018-11-30'].reset_index().copy()
val = data[data.obs_mth == '2018-11-30'].reset_index().copy()
#这是我们全部的变量，info结尾的是自己做的无监督系统输出的个人表现，score结尾的是收费的外部征信数据
lst = ['person_info','finance_info','credit_info','act_info','td_score','jxl_score','mj_score','rh_score']
df_train = df_train.sort_values(by = 'obs_mth',ascending = False)
df_train.head()
df_train = df_train.sort_values(by = 'obs_mth',ascending = False)
rank_lst = []
for i in range(1,len(df_train)+1):
rank_lst.append(i)
df_train['rank'] = rank_lst
df_train['rank'] = df_train['rank']/len(df_train)
pct_lst = []
for x in df_train['rank']:
if x <= 0.2:
x = 1
elif x <= 0.4:
x = 2
elif x <= 0.6:
x = 3
elif x <= 0.8:
x = 4
else:
x = 5
pct_lst.append(x)
df_train['rank'] = pct_lst
#train = train.drop('obs_mth',axis = 1)
df_train.head()
df_train['rank'].groupby(df_train['rank']).count()
#rank
#1 15966
#2 15966
#3 15966
#4 15966
#5 15967
#Name: rank, dtype: int64
#定义lgb函数
def LGB_test(train_x,train_y,test_x,test_y):
from multiprocessing import cpu_count
clf = lgb.LGBMClassifier(
boosting_type='gbdt', num_leaves=31, reg_alpha=0.0, reg_lambda=1,
max_depth=2, n_estimators=800, objective='binary',
subsample=0.7, colsample_bytree=0.7, subsample_freq=1,
learning_rate=0.05, min_child_weight=50,random_state=None,n_jobs=cpu_count()-1,
num_iterations = 800 #迭代次数
)
clf.fit(train_x, train_y,eval_set=[(train_x, train_y),(test_x,test_y)],eval_metric='auc',early_stopping_rounds=100)
print(clf.n_features_)
return clf,clf.best_score_[ 'valid_1']['auc']
feature_lst = {}
ks_train_lst = []
ks_test_lst = []
for rk in set(df_train['rank']):
# 测试集8.18以后作为跨时间验证集
#定义模型训练集与测试集
ttest = df_train[df_train['rank'] == rk]
ttrain = df_train[df_train['rank'] != rk]
train = ttrain[lst]
train_y = ttrain.bad_ind
test = ttest[lst]
test_y = ttest.bad_ind
start = time.time()
model,auc = LGB_test(train,train_y,test,test_y)
end = time.time()
#模型贡献度放在feture中
feature = pd.DataFrame(
{'name' : model.booster_.feature_name(),
'importance' : model.feature_importances_
}).sort_values(by = ['importance'],ascending = False)
#计算训练集、测试集、验证集上的KS和AUC
y_pred_train_lgb = model.predict_proba(train)[:, 1]
y_pred_test_lgb = model.predict_proba(test)[:, 1]
train_fpr_lgb, train_tpr_lgb, _ = roc_curve(train_y, y_pred_train_lgb)
test_fpr_lgb, test_tpr_lgb, _ = roc_curve(test_y, y_pred_test_lgb)
train_ks = abs(train_fpr_lgb - train_tpr_lgb).max()
test_ks = abs(test_fpr_lgb - test_tpr_lgb).max()
train_auc = metrics.auc(train_fpr_lgb, train_tpr_lgb)
test_auc = metrics.auc(test_fpr_lgb, test_tpr_lgb)
ks_train_lst.append(train_ks)
ks_test_lst.append(test_ks)
feature_lst[str(rk)] = feature[feature.importance>=20].name
train_ks = np.mean(ks_train_lst)
test_ks = np.mean(ks_test_lst)
ft_lst = {}
for i in range(1,6):
ft_lst[str(i)] = feature_lst[str(i)]
fn_lst=list(set(ft_lst['1']) & set(ft_lst['2'])
& set(ft_lst['3']) & set(ft_lst['4']) &set(ft_lst['5']))
print('train_ks: ',train_ks)
print('test_ks: ',test_ks)
print('ft_lst: ',fn_lst )
#[LightGBM] [Warning] Unknown parameter: max_features
#[1] training's auc: 0.726731 training's binary_logloss: 0.0827979 valid_1's auc: 0.742666 valid_1's binary_logloss: 0.12066
#[2] training's auc: 0.769499 training's binary_logloss: 0.0822062 valid_1's auc: 0.753919 valid_1's binary_logloss: 0.119728
#[3] training's auc: 0.788952 training's binary_logloss: 0.0816227 valid_1's auc: 0.762911 valid_1's binary_logloss: 0.118777
#. . .
#[188] training's auc: 0.827082 training's binary_logloss: 0.0777181 valid_1's auc: 0.786679 valid_1's binary_logloss: 0.078782
#[189] training's auc: 0.827128 training's binary_logloss: 0.0777136 valid_1's auc: 0.786756 valid_1's binary_logloss: 0.0787781
#[190] training's auc: 0.827162 training's binary_logloss: 0.0777108 valid_1's auc: 0.786696 valid_1's binary_logloss: 0.0787811
#train_ks: 0.4907124806547195
#test_ks: 0.47382530047645305
#ft_lst: ['credit_info', 'person_info', 'finance_info']
lst = ['person_info','finance_info','credit_info','act_info']
train = data[data.obs_mth != '2018-11-30'].reset_index().copy()
evl = data[data.obs_mth == '2018-11-30'].reset_index().copy()
x = train[lst]
y = train['bad_ind']
evl_x = evl[lst]
evl_y = evl['bad_ind']
model,auc = LGB_test(x,y,evl_x,evl_y)
y_pred = model.predict_proba(x)[:,1]
fpr_lgb_train,tpr_lgb_train,_ = roc_curve(y,y_pred)
train_ks = abs(fpr_lgb_train - tpr_lgb_train).max()
print('train_ks : ',train_ks)
y_pred = model.predict_proba(evl_x)[:,1]
fpr_lgb,tpr_lgb,_ = roc_curve(evl_y,y_pred)
evl_ks = abs(fpr_lgb - tpr_lgb).max()
print('evl_ks : ',evl_ks)
from matplotlib import pyplot as plt
plt.plot(fpr_lgb_train,tpr_lgb_train,label = 'train LR')
plt.plot(fpr_lgb,tpr_lgb,label = 'evl LR')
plt.plot([0,1],[0,1],'k--')
plt.xlabel('False positive rate')
plt.ylabel('True positive rate')
plt.title('ROC Curve')
plt.legend(loc = 'best')
plt.show()
#[1] training's binary_logloss: 0.090317 training's auc: 0.712883 valid_1's binary_logloss: 0.0986629 valid_1's auc: 0.678619
#Training until validation scores don't improve for 100 rounds.
#[2] training's binary_logloss: 0.0896369 training's auc: 0.779216 valid_1's binary_logloss: 0.0978883 valid_1's auc: 0.755811
#[3] training's binary_logloss: 0.0885026 training's auc: 0.779149 valid_1's binary_logloss: 0.0966811 valid_1's auc: 0.749375
#[4] training's binary_logloss: 0.087998 training's auc: 0.780539 valid_1's binary_logloss: 0.0961527 valid_1's auc: 0.759009
#...
#[179] training's binary_logloss: 0.0784288 training's auc: 0.812571 valid_1's binary_logloss: 0.0900886 valid_1's auc: 0.779962
#[180] training's binary_logloss: 0.0784267 training's auc: 0.812602 valid_1's binary_logloss: 0.0900914 valid_1's auc: 0.779887
#[181] training's binary_logloss: 0.078425 training's auc: 0.812601 valid_1's binary_logloss: 0.0900941 valid_1's auc: 0.779927
#[182] training's binary_logloss: 0.0784229 training's auc: 0.8126 valid_1's binary_logloss: 0.0900964 valid_1's auc: 0.779932
#Early stopping, best iteration is:
#[82] training's binary_logloss: 0.0788374 training's auc: 0.811646 valid_1's binary_logloss: 0.089958 valid_1's auc: 0.779946
#4
#train_ks : 0.4801091876625077
#evl_ks : 0.4416674980164514
LightGBM其实效果确实是比较LR要好的，但是我们LR也可以逼近这个效果，下节课我们会具体来做。
评分卡公式变形
600
+
50
×
ln
P
0
P
1
ln
2
,
P
0
为
好
人
，
P
1
为
坏
人
600+50 \times \frac{\ln \frac{P_{0}}{P_{1}}}{\ln 2},P_{0}为好人，P_{1}为坏人
600+50×ln2lnP1P0,P0为好人，P1为坏人
600
+
50
×
ln
1
−
x
b
e
t
a
x
b
e
t
a
ln
2
600+50 \times \frac{\ln \frac{1-xbeta}{xbeta}}{\ln 2}
600+50×ln2lnxbeta1−xbeta
600
+
50
×
log
2
1
−
x
b
e
t
a
x
b
e
t
a
600+50 \times \log _{2} \frac{1-{ xbeta }}{{ xbeta }}
600+50×log2xbeta1−xbeta
#['person_info','finance_info','credit_info','act_info']
#算分数onekey
def score(xbeta):
score = 1000+500*(math.log2(1-xbeta)/xbeta) #好人的概率/坏人的概率
return score
evl['xbeta'] = model.predict_proba(evl_x)[:,1]
evl['score'] = evl.apply(lambda x : score(x.xbeta) ,axis=1)
fpr_lr,tpr_lr,_ = roc_curve(evl_y,evl['score'])
evl_ks = abs(fpr_lr - tpr_lr).max()
print('val_ks : ',evl_ks)
#val_ks : 0.4416674980164514
#生成报告
row_num, col_num = 0, 0
bins = 20
Y_predict = evl['xbeta']
Y = evl_y
nrows = Y.shape[0]
lis = [(Y_predict[i], Y[i]) for i in range(nrows)]
ks_lis = sorted(lis, key=lambda x: x[0], reverse=True)
bin_num = int(nrows/bins+1)
bad = sum([1 for (p, y) in ks_lis if y > 0.5])
good = sum([1 for (p, y) in ks_lis if y <= 0.5])
bad_cnt, good_cnt = 0, 0
KS = []
BAD = []
GOOD = []
BAD_CNT = []
GOOD_CNT = []
BAD_PCTG = []
BADRATE = []
dct_report = {}
for j in range(bins):
ds = ks_lis[j*bin_num: min((j+1)*bin_num, nrows)]
bad1 = sum([1 for (p, y) in ds if y > 0.5])
good1 = sum([1 for (p, y) in ds if y <= 0.5])
bad_cnt += bad1
good_cnt += good1
bad_pctg = round(bad_cnt/sum(evl_y),3)
badrate = round(bad1/(bad1+good1),3)
ks = round(math.fabs((bad_cnt / bad) - (good_cnt / good)),3)
KS.append(ks)
BAD.append(bad1)
GOOD.append(good1)
BAD_CNT.append(bad_cnt)
GOOD_CNT.append(good_cnt)
BAD_PCTG.append(bad_pctg)
BADRATE.append(badrate)
dct_report['KS'] = KS
dct_report['BAD'] = BAD
dct_report['GOOD'] = GOOD
dct_report['BAD_CNT'] = BAD_CNT
dct_report['GOOD_CNT'] = GOOD_CNT
dct_report['BAD_PCTG'] = BAD_PCTG
dct_report['BADRATE'] = BADRATE
val_repot = pd.DataFrame(dct_report)
val_repot
扫一扫
721
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
