A leading German Bank wants to identify their customers who are likely to fraudulent through their historical information of Fraudsters.
Demographic data like Gender, Marital status, and Job level and transactional data likes.
Average monthly balance maintained in savings bank account and their purpose of the loan was found to be influencing.
Identification of Fraudulent customers
Random Forest Model was built for active and non-active customers by assigning the Status of being Fraudulent as Dependent variable and all other attributes as Independent variable
The accuracy of the model was assessed by checking the model results against the 20% customer base (Control Group)
Most of the customers being Fraudulent are the customers who have taken vehicle loan mostly car loan.
Based on the analysis, it was observed that single customers who are skilled employees and who maintain an average monthly balance of less than 100 DM (Deutsche Mark) are most likely to be Fraudulent.
15 Percent of the customers were identified as potential Fraudsters from the current active customer base and the same has been communicated to their Credit Team.