Optimizing Inventory Management In Retail With Hybrid Genetic Algorithms And LSTM

Abstract

In the fast-changing retail industry, maintaining a balance between reducing operating expenses and meeting service demands is essential to effective inventory management. The traditional models are not as efficient as resolving the non-linear behavior of contemporary consumer demands. In this work, a Hybrid GA-LSTM model is proposed as an integrated model that utilizes an LSTM neural network for high-accuracy demand forecasting and Genetic Algorithms (GA) for multi-objective inventory optimization. The proposed model was compared to the conventional models, such as Economic Order Quantity (EOQ) and standalone LSTM models. The empirical result shows that the Hybrid GA-LSTM approach can effectively improve the performance of these benchmarks by reducing total inventory costs by 37.53%, maintaining 95% service level, and minimizing the stockout rate to 5%. The results indicate that using machine learning to improve the prediction and evolutionary algorithms for decision-making can better match the inventory to the real market. In practice, this study offers a flexible and efficient data-driven approach for retailers to deal with seasonality and promotions. Although it may have limitations in terms of computational requirements and the need for a vast database of historical data, the model could be expanded to incorporate more data and other factors, thereby improving its accuracy and applicability as time goes on. The study is important to the field because it shows the efficacy of the hybrid optimization, and it has implications beyond the narrow scope of the study, such as the healthcare and manufacturing industries. To further validate the effectiveness of the long-term solution, future studies could increase scalability and investigate the integration of other optimization methods, like reinforcement learning, in real operational settings.