CostGuard: Browser-Side Machine Learning Framework for Multi-Platform Price Trend Prediction

Abstract

Dynamic price fluctuations across multi, vendor online retail platforms make timely purchasing decisions difficult for consumers. Most traditional price, tracking systems rely on historical analysis or alert users when prices cross predefined thresholds, but they do not predict short, term future price changes. Cloud, centric machine learning approaches for price prediction increase infrastructure cost, reduce scalability, introduce privacy concerns, and cause interaction latency. CostGuard AI presents a decentralized, browser, executed machine learning framework that performs short, term price prediction locally using TensorFlow.js. A synthetic dataset is generated to mimic realistic e, commerce pricing patterns while remaining compliant with platform usage policies. A sliding, window neural regression model predicts the next, interval minimum price across multiple platforms. Experimental results show acceptable prediction accuracy, very low inference latency, and negligible backend resource utilization, demonstrating the feasibility of scalable and privacy, preserving browser, based price forecasting systems.