Integrated ML and Big Data Approach for Airline Chargeback Fraud Detection

Airline ticket booking using credit cards has become an important mode of transaction, occupying more than 60% of total transactions. However, the chargeback fraud associated with it is also growing significantly along with the widespread increase in the number of total credit card transactions. This vulnerability, many times, goes undetected until the fraudulent transaction is investigated and confirmed. By the time it reaches the confirmation stage, Airlines need to spend a lot of time on investigation and there is a substantial risk of losing significant money, as it is not a recoverable loss. In addition, fraudsters quickly adapt to other methods, which cannot be easily identifiable unless a comprehensive investigation happens. Given the continuously changing nature of fraud transactions, Machine Learning Models could be of significant use in predicting fraudulent transactions. Many Machine Learning and Deep Learning approaches have evolved in detecting various credit card fraudulent transactions. However, a Machine Learning approach with a single technique may fail to capture the multidimensional aspect of fraudulent transactions. Hence, there is a need for a comprehensive integrated approach, which captures the deviations in trend, pattern, anomalies at a given time. On top of the ML solution, integration with the optimally designed Big Data platform makes the solution more robust in predicting fraudulent transactions in real time. This paper discusses the problem statement, business case, solution approach, empirical analysis, modelling, integration & implementation, and results.