Research on E-Commerce Product Inventory Monitoring and Sales Forecasting Based on the Spark Platform: A Case Study of Furniture Products

This study presents a forecasting and inventory management framework for e-commerce, integrating Long Short-Term Memory (LSTM) networks and Extreme Gradient Boosting (XGBoost) within a distributed computing environment using Apache Spark. The approach is tested on the UK Online Retail dataset for a specific furniture SKU. Key contributions include integrating deep learning and ensemble learning, implementing a scalable Spark-based pipeline, and linking forecasts directly to inventory replenishment strategies. Experimental results show complementary strengths: LSTM excels at capturing long-term trends, while XGBoost responds quickly to short-term fluctuations. The framework demonstrates practical potential for forecast-driven inventory optimization.