Application of Classification Models on Fraud Detection in Retail

Data-based analytical approach in this area is becoming increasingly important for identification of new methods and approaches helping retail owners protect themselves from financial losses while maintaining customer confidence. The number of self-checkout stations in various retail stores is constantly increasing. This automated process helps avoid long lines and speeds up the paying process for individual customers. On the other hand, a certain level of trust with the customer is required by the retailers, supported by personnel checking the payment from the handheld mobile scanner. This situation requires a good strategy for how to plan enough necessary checks and this number is not annoying for customers. The most important objective is to identify as many false scans as possible. Based on the public data available from Data Mining Cup we have generated a batch of different classification models to find an optimal decision support model from a defined business perspective.