I am working for a logistics firm and there are approx. 750+ customers who avail our services. I am in the process of building and generating some insight for the business based on the payments made by these customers for the last 1 year. Some make payment on time whereas some are late & some are extremely late. Could you please advice which modeling technique or statistical approach would be best in this case. I can think of creating clusters based on the payment history and highlights & placing all defaulters in one cluster where our company can focus. PLease suggest.
2
-
1$\begingroup$ "Generate insight" is far too broad; until you know what you're asking it's not useful to ask about tools. $\endgroup$– Sean OwenApr 10, 2015 at 21:33
-
$\begingroup$ Hi Sean, Generating insight by here I mean- finding insights such as certain descriptive statistics- mean payment days, median payment days, default by how many days, creating a delay band i.e. 1-10 days, 11-20 days etc. region wise details, BU wise details. $\endgroup$– ShiviApr 11, 2015 at 5:02
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Here are some things that come to mind:
1) ARMA/ARIMA models: These may show if there is any seasonal trends in their payment history.
2) Logistic Regression: Will allow you model the conditional probability of a missed payments based off of number of previous missed payments.
From here you need to determine what the threshold for risk is since that's a little more subjective.
