A Computational Optimization Framework for Dynamic Pricing in E-Commerce Using Integrated Forecasting and Learning Algorithms

Abstract This paper proposes a computational optimization framework for dynamic pricing in e-commerce by integrating time-series forecasting and adaptive learning mechanisms within a constrained stochastic optimization model. Demand is estimated using an ARIMA-based forecasting module and incorporated into a rolling-horizon revenue maximization problem under inventory and price constraints. A gradient-based adaptive update rule dynamically adjusts prices in response to observed demand. Closed-form optimality conditions are derived, convergence properties are established, and computational complexity is analyzed. Sensitivity analysis demonstrates robustness with respect to elasticity, demand uncertainty, and inventory levels. The framework offers a scalable computational solution for real-time revenue management.