Logistic Regression-based Classification of Inventory Control Systems: A Comparative Study of Q-system and P-system in Batik Retail Products

Abstract

Inventory control plays a pivotal role in ensuring product availability and operational efficiency, especially within dynamic retail environments such as the Indonesian batik industry.

Traditional systems like the Fixed Order Quantity (Q-System) and Fixed Order Period (P- System) are often selected based on managerial judgment, which may lead to suboptimal

decisions due to the variability in demand and product characteristics. This study introduces a data-driven approach to classify inventory items into Q-System or P-System categories using logistic regression. Historical inventory data from a batik retail business in Pekalongan, encompassing demand, lead time, cost, and profitability for 15 high-performing products, was used to build and train the model. Each product's inventory control strategy was first evaluated based on total cost minimization, combining ordering cost, holding cost, and lost profit. The best strategy (Q or P) per product was determined and used as a classification label. Logistic regression was then applied, using normalized statistical and operational variables, to develop a predictive model. The model achieved perfect classification performance on training and testing datasets, with a Mean Squared Error (MSE) of 0, demonstrating its effectiveness. This approach enhances decision-making transparency, optimizes inventory costs, and bridges traditional inventory theory with modern predictive analytics. The findings offer practical implications for retailers aiming to adopt interpretable, data-informed inventory strategies.