Product Return Prediction in E-Commerce Platforms

Abstract

Online shopping is fast and easy,but returning items is just as easy, and that’s becoming a big problem for online stores. Every return wastes time, money, and effort. This project brings together smart ideas from recent research to build a system that can predict returns even before a customer places an order. Many systems already use things like past purchases, product details, or customer reviews to guess return chances. But we’re adding something new: Digital Touch Latency (DTL) a way to track how long a shopper hesitates, like checking the size chart or switching colors. These small actions often mean the customer is unsure, and unsure customers are more likely to return things.We’re also using a Sentiment Drift Engine to watch how product reviews change over time. If people start giving worse reviews suddenly, our system notices and increases the return risk right away. By combining hesitation behavior (DTL), review trends, and regular shopping data, our machine learning model can make much better predictions. This helps stores act early by showing better recommendations, fixing product info, or warning about risky products.In short, fewer returns help stores spend less and keep customers happy with their shopping.