Adapting AI Models for Dynamic Inventory Optimization in E-Commerce

Abstract

This study investigates how Artificial Intelligence (AI) technologies-specifically Machine Learning (ML), Deep Learning (DL), and Transformer-based models including Large Language Models (LLMs)-can be effectively adapted to optimize inventory management in e-commerce environments. The research addresses the question: How can different AI methodologies be integrated to minimize stock inefficiencies while maximizing operational efficiency in real-time e-commerce environments? Using a structured bibliometric and bibliographic analysis methodology of 163 scholarly articles from the Scopus database, this study identifies and evaluates implementation patterns of AI models in inventory optimization. Results demonstrate that ML-based models like decision trees and random forests achieve up to 98% accuracy in demand forecasting by analyzing historical data, while DL models such as LSTM networks excel at time-series analysis for enhanced stock predictions. Transformer models and LLMs leverage unstructured data to provide real-time insights into consumer sentiment, enabling more agile inventory adjustments. This research contributes to e-commerce operations management by providing a framework for selecting appropriate AI methodologies based on specific inventory challenges, with implications for minimizing costs and improving customer satisfaction through optimized stock management.