Hybrid EEMD-LSTM Model for Accurate E-commerce Demand Forecasting

Abstract

Accurate demand forecasting in e-commerce is essential for optimizing inventory management, reducing overhead costs, and enhancing customer satisfaction. However, traditional forecasting methods often fall short in handling the nonlinearity and noise inherent in e-commerce sales data. This research proposes a novel hybrid model combining Ensemble Empirical Mode Decomposition (EEMD) with Long Short-Term Memory (LSTM) networks to address these challenges. EEMD effectively decomposes the original sales time series into a set of intrinsic mode functions (IMFs), capturing both high-frequency and low-frequency components. These decomposed signals are then individually modeled using LSTM networks, which are adept at capturing temporal dependencies and learning from sequential data. The proposed hybrid EEMD-LSTM model is validated using real-world e-commerce datasets, and its performance is benchmarked against traditional time series models and standalone LSTM models. The results demonstrate a significant improvement in forecasting accuracy, showcasing the potential of hybrid deep learning frameworks in e-commerce analytics.