Pharmaceutical Inventory Optimization through Predictive Machine Learning and Business Intelligence Techniques

Abstract

Inadequate pharmaceutical inventory management generates stockouts in 69% of establishments and economic losses of 3.5-8.3% of annual revenue, affecting the continuity of medical treatments. This systematic review analyzes the effectiveness of Machine Learning and Business Intelligence techniques to optimize demand forecasting in small and medium pharmacies with limited technology. PRISMA methodology was implemented to identify relevant studies in Scopus between 2020-2025, using structured PICOC criteria. From 1,327 initial records, 40 studies that met specific inclusion criteria for pharmaceutical predictive analysis were selected. Data was extracted through standardized protocol, evaluating predictive accuracy, operational impact, and implementation barriers. Results demonstrate significant improvements of 30.5% in predictive accuracy compared to traditional methods. Neural networks showed highest effectiveness with 85.3% average accuracy, followed by ensemble models with 83.7%. A 28.5% reduction in stockouts and 25.7% inventory optimization was achieved. Main barriers identified include technical training limitations in 58% of staff and inadequate infrastructure in 43% of establishments. Machine Learning and Business Intelligence techniques represent scalable solutions to optimize pharmaceutical inventories, being viable even in resource-limited contexts, significantly contributing to economic sustainability and public health improvement.