03-06-2017 03:17 PM
Hello, Fellow Users,
I am building an attrition model. I am ending up with 10 predictors. The client wants to know for each record, which the top reason it will attride (like the top reasons the credit card application is denied).
I haven't worked in credit card industry, so I don't know exactly how they do it. One solution I am thinking is:
1), Standardize the predictors
2), Estimate the coefficient
3), Scoring each record
4), Rank the absolute value of coefficient * preditors
Because the end value is due to the summary of each coefficient * preditors, I think I am ranking the important of each predictor.
Could anyone give me some insight whether this works?
Thanks a lot.