Hybrid Machine Learning Models for Optimizing Retail Market and Inventory Forecasting

Abstract

Effective inventory management is critical for retail operations, relying heavily on accurate sales data analysis to optimize stock levels, forecast demand, and minimize supply chain inefficiencies. Traditional machine learning models such as Random Forest (RF), K-Nearest Neighbors (KNN), Logistic Regression (LR), Multinomial Naïve Bayes (MNB), and Support Vector Machine (SVM) have been used to classify sales data, but their limitations in handling complex retail datasets often result in suboptimal performance. In this regard, this research proposed a novel hybrid model based on K-Nearest Neighbors and Support Vector Machine that strategically combines the capabilities of KNN's in local pattern recognition with SVM's in high-dimensional classification for optimizing retail market and inventory forecasting. Through this research on real-world sales data, the proposed hybrid model shows superior performance, by achiving accuracy of 0.9999, precision, 0.9808 recall, 0.9808 F1-score, and 0.9991 accuracy showing enhanced performance in comparison to the baseline models. These findings support the effectiveness of hybrid machine learning algorithms in retail analytics, which provide significant gains in sales classification accuracy and inventory prediction reliability. As a result, retailers can make more informed stocking decisions, reduce waste, and enhance overall operational efficiency through data-driven insights.