如何将逻辑回归模型转换成加法评分卡,比如下面的套现分析中,变量LIMITUSAGE取值范围多少到多少,加分还是减少分
> TX.fit <- glm(TX_IND~
+ +LIMITUSAGE # 额度使用率
+ +RES21 # 交易商户数量
+ # +LAST_24MON_OVD_TERM_1
+ +LAST_24MON_OVD_TERM_19 #交易商户数量
+ +TOP_MERCHANT_PER # 最高交易商户交易金额占比
+ +GOOD_PER #非低零扣率交易金额占比
+ +CASHRATIO_M1
+ ,family = binomial(link='logit'),data=TX_8479_M1_DAT);
Warning message:
glm.fit: fitted probabilities numerically 0 or 1 occurred
> summary(TX.fit)
Call:
glm(formula = TX_IND ~ +LIMITUSAGE + RES21 + LAST_24MON_OVD_TERM_19 +
TOP_MERCHANT_PER + GOOD_PER + CASHRATIO_M1, family = binomial(link = "logit"),
data = TX_8479_M1_DAT)
Deviance Residuals:
Min 1Q Median 3Q Max
-8.4904 -0.0012 0.1961 0.2716 6.3373
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.537e+01 4.408e-01 -34.87 <2e-16 ***
LIMITUSAGE 8.968e-01 4.520e-02 19.84 <2e-16 ***
RES21 2.051e-01 4.787e-03 42.85 <2e-16 ***
LAST_24MON_OVD_TERM_19 -3.656e-01 6.615e-03 -55.26 <2e-16 ***
TOP_MERCHANT_PER 1.928e+01 4.445e-01 43.38 <2e-16 ***
GOOD_PER -6.555e+00 3.801e-02 -172.46 <2e-16 ***
CASHRATIO_M1 -1.883e-02 3.133e-04 -60.09 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 179030 on 156539 degrees of freedom
Residual deviance: 61874 on 156533 degrees of freedom
AIC: 61888
Number of Fisher Scoring iterations: 7