Inventory management is one of the most important elements of supply chain management. This study explores the critical challenges in retail inventory management, particularly the discrepancies between physical inventory and records in corporate databases, and introduces a novel machine learning-based approach for enhancing inventory accuracy. Through comprehensive data analysis from a retail company in Turkey, the study applies classification algorithms such as Support Vector Machine (SVM), Decision Trees, Extreme Gradient Boosting (XGBoost), and Random Forest, optimized by employing feature selection with Particle Swarm Optimization, to predict and detect inventory inaccuracies effectively. The research culminates in the development of a decision support system designed to assist store personnel in identifying potential stock discrepancies, thereby facilitating proactive inventory management and enhancing store performance and customer satisfaction. This study discusses the methodology and results of implementing these algorithms, the integration of a sophisticated feature selection process, and the effectiveness of the proposed system in real-world retail settings. Ultimately, this study not only addresses prevalent issues in inventory management but also contributes to the evolving field of supply chain analytics by demonstrating how machine learning can significantly improve the accuracy of inventory records.