AI Driven Inventory Optimization Framework Using Deep Learning and Metaheuristic Algorithms

Optimized stock management is a crucial step to supply chain improvement, reducing operation costs, and providing ‎high Service levels. In this study, a New model using Simulated Annealing (SA) and a hybrid optimization approach is ‎put forward for optimizing supply chain stock process. The SA algorithm continuously enhances inventory solutions ‎endeavoring to find near-optimal setups that balance service levels and customers' requirements. First, the supply chain ‎dataset was collected and preprocessed with K-Nearest Neighbour (KNN) imputation for handling missing values and z-‎score normalization for scaling. For handling class imbalance, data augmentation was carried out using Synthetic ‎Minority Over-sampling Technique (SMOTE). Subsequently, Deep Maxout Network (DMN) was utilized to perform ‎feature extraction in order to identify key patterns and relationships within the data. The inventory optimization ‎problem was subsequently tackled using the SA algorithm enhanced by a hybrid optimization approach (SCTDO) that ‎integrates the exploratory ability of the Sine Cosine Algorithm (SCA) and the refinement ability of the Tasmanian Devil ‎Optimization (TDO) algorithm. The hybrid approach is robust at efficient exploration and exploitation of the solution ‎space. The proposed model was implemented in Python and evaluated on different performance criteria like correlation ‎coefficient, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and computation time. The proposed ‎algorithm produced an RMSE value of 0.017, MAE of 0.013, computation time of 0.93, and a correlation coefficient of ‎‎0.983. Comparative study with other models also reflected the improved performance of the proposed Framework.