Fraud Prediction in Pakistani E-commerce Market

With Pakistan being ranked as the 46th largest revenue generator in terms of the E-commerce industry, online frauds have increased proportionally. The process of online shopping has changed drastically as the seller and buyer can now communicate directly through social media applications without needing a specific platform. It implies that all fraud prevention techniques, already in place, fail in such scenarios as they are only applicable to their platform. So, for an easily attainable input to the fraud prevention pipeline, our research focuses on analyzing the fraudulent activities in this market by using commonly available customer, product, and seller traits, as features. For this research, a product-based fraud detection dataset was collected through a survey and various feature selection techniques and ML models were applied to it. In prospect, our approach can be used to develop a utility that automatically extracts relevant features, calculates risk scores, and facilitates customers in purchase decisions, given a threshold on the risk score.