April 10, 2015
Experiment to get inputs for Prescriptive Analytics on Consumer complaints for financial products.
Classification Algorithm - Two Class Boosted Decision Tree wins over Two Class Logistics Regression with AUC as 0.876 over 0.872. The web service derived from Azure ML could be integrated with Complaint logging system so that support executive could get the probability of whether the issue would be “Closed with monetary relief”. If yes, the company could take that issue as critical and start with balance maintenance and aim for non-monetary issue resolution.\n