Online Shopping Website Review System Using Predictive AI

Abstract— This document introduces a novel AI-driven framework for validating online shopping reviews by integrating a BERT-based semantic analyzer with XGBoost behavioral modeling. The system allows customers to submit natural- language feedback, which is automatically classified as genuine or fraudulent in real time. Classified outputs update the review database: authentic entries are published immediately, while suspect ones trigger automated moderation actions (account verification, vendor alerts). Designed for non-technical stakeholders, it provides clear, human-readable rationales for each decision. Applications include e-commerce platforms, customer support dashboards, and merchant analytics tools. Experimental evaluation on a multi-category retail dataset demonstrates 93 % classification accuracy and a 38 % drop in fraudulent accounts, underscoring its potential to enhance trust without manual oversight.