Optimizing Inventory Management and Demand Forecasting with LSTM Neural Networks and Machine Learning: An Integrated Approach with ABC-DEA Classification

Abstract

This research project takes a comprehensive approach to e-commerce inventory management by incorporating modern machine learning and optimization approaches. We provide efficient inventory strategies by utilizing LSTM neural networks for demand forecasting, FAHP-DEA techniques for ABC product classification, and a modified DDMR-based expiration risk assessment with Q learning. We evaluated several machine learning classifiers, including Random Forest, Decision Tree, SVM, Naïve Bayes, and KNN. Random Forest was the most accurate (97.2%). In addition, we introduce Q-learning as an optimization technique cascaded with expiry risk assessment to refine inventory strategies and respond to market changes in real time. This iterative approach provides appropriate inventory levels, efficient operations, and increased customer satisfaction. In essence, our methodology offers actionable information for e-commerce platforms to optimize inventory processes.