A Hybrid Demand Forecasting and Adaptive Optimization Approach for Dynamic Pricing in E-Commerce

Small and mid-size e-commerce sellers lose profit because automated pricing tools react to competitors without understanding customer demand. This paper proposes a pricing system that predicts demand, analyses price sensitivity, and recommends optimal prices in real time. Tested on real retail transaction data from the UCI Online Retail II dataset, the system achieves 11.8% mean profit improvement. Further experiments on 80 consumer electronics products show 27.1% profit improvement over standard pricing, with significantly more stable recommendations than existing approaches. The complete system delivers pricing decisions in under 120 milliseconds, making it practical for everyday seller use. Code and data are publicly available at https://doi.org/10.5281/zenodo.19151881.