Developing predictive analytics for ABC-VEN matrix inventory management in Kenyan hospitals

Abstract

Hospitals and healthcare facilities depend on robust inventory management systems to ensure that medications are available to patients promptly while keeping operational costs in check. Traditional inventory methods often fall short when it comes to adjusting for factors such as fluctuating demand, seasonal trends, and external disruptions. These shortcomings can lead to inefficiencies in determining restocking points and managing stock levels. This study addresses these problems using the CRISP-ML(Q) methodology and in modeling, the power of machine learning (ML) algorithms and Mathematical optimization within the ABC-VEN matrix framework. The aim of the research is to accurately forecast the restock quantities and timing of pharmaceutical supplies by means of stochastic inventory optimization models based on Reorder Point (ROP) analysis. Past sales data, along with the ABC-VEN classification, was used to build a predictive model to estimate key inventory metrics. The hybrid approach then computed figures such as mean demand (μD), demand variability (σD), safety stock (SS), reorder point (ROP), ideal order size, stock levels, and total inventory cost. Finally, ROP forecasts were produced using these outcomes. The model’s accuracy was validated using the coefficient of determination (R2), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). With an R2 score close to 0.85, the model proved to be highly effective, accounting for 85% of the variance in reorder point predictions. This performance underlines the model’s capability in enhancing inventory control. Ultimately, the optimization model helped reduce both stockouts and overstocking by refining reorder point strategies and cutting down unnecessary expenses. This research highlights the valuable role that ML and optimization can play in strengthening hospital pharmacy operations while maintaining low costs and reliable drug availability. KEY WORDS: Stockouts, Overstocking, EOQ, ROP, Machine Learning, Optimization, Model Performance