Leveraging LSTM and Prophet Models for Enhanced AI-Driven Demand Prediction in E-Commerce

Abstract

This research paper explores the integration of Long Short-Term Memory (LSTM) networks and Prophet models to enhance demand prediction accuracy in the e-commerce sector, addressing the dynamic and complex nature of consumer behavior. The study evaluates the performance of these models using historical sales data from several large e-commerce platforms, focusing on key metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). LSTM, known for its ability to capture long-term dependencies in sequential data, is combined with Prophet, which excels in handling time series data with seasonal trends and missing values. Our hybrid approach leverages the strengths of LSTM in recognizing patterns and Prophet’s capability to incorporate seasonality and holiday effects, resulting in predictions that outperform traditional time series models. Extensive experiments demonstrate that the proposed model achieves a significant reduction in prediction error compared to stand-alone models, with a 15-20% improvement in accuracy. Furthermore, the paper discusses the computational efficiency of the hybrid model, highlighting its scalability for large datasets typical of e-commerce environments. The findings suggest that integrating LSTM and Prophet models provides a robust framework for real-time demand forecasting, offering valuable insights for inventory management and strategic decision-making. This study contributes to the existing literature by demonstrating the potential of combining machine learning techniques for practical applications in demand prediction, paving the way for future research in hybrid modeling approaches.