ADAPTIVE PRICING PREDICTION BASED ON USER INTENT SIGNALS

Abstract

Adaptive pricing has become an essential component of modern e-commerce platforms as businesses attempt to respond to rapidly changing consumer behaviour and competitive market conditions. Traditional pricing strategies such as fixed pricing, rule-based discounting, and inventory-driven dynamic pricing are limited because they primarily rely on historical data and static business rules rather than real-time user behaviour. This research proposes an Adaptive Pricing Prediction System based on User Intent Signals to enhance pricing strategies by analysing behavioural indicators generated during user interactions on an e-commerce platform. The system collects behavioural signals such as product revisit frequency, browsing duration, click patterns, and exit-reentry behaviour to infer a customer’s purchase intent and price sensitivity. These signals are processed through feature engineering techniques to generate meaningful numerical representations suitable for machine learning models. A Light Gradient Boosting Machine (LightGBM) regression model is used to predict personalized discount percentages for each user session and product combination. The predicted discount is dynamically applied to the base price to generate an optimized adaptive price in real time. The system architecture integrates a React-based frontend interface, a Java Spring Boot backend for authentication and API services, and a Pythonbased machine learning module for predictive modelling. Experimental results indicate that incorporating user intent signals significantly improves pricing accuracy and conversion potential compared to traditional rule-based pricing approaches. The proposed framework demonstrates how behavioural analytics and machine learning can be integrated to create intelligent pricing mechanisms that improve revenue optimization, customer engagement, and personalized shopping experiences in digital commerce platforms.