A Machine Learning-Based Stock Forecasting Method for Inventory Optimization in Micro and Small Enterprises

Efficient inventory management remains a critical challenge for Micro and Small Enterprises (MSEs) that operate under limited resources and fluctuating market demands. This study proposes a lightweight and interpretable machine-learning framework based on the Random Forest algorithm to predict product demand and optimize inventory levels. Historical sales data was preprocessed, structured, and used to train and validate the model through multiple evaluation metrics. The proposed model achieved a Mean Absolute Percentage Error (MAPE) of 2.41% and a Coefficient of Determination (R²) of 0.99, outperforming comparative models such as K-Nearest Neighbors, Decision Tree, and XGBoost. These results confirm the model's capacity to capture short-term fluctuations and long-term trends with high predictive accuracy. Feature-importance analysis revealed that the interaction between quantity and price was the most influential variable, followed by relative price and seasonal factors. The findings demonstrate that data-driven forecasting can significantly reduce overstocking and stockout situations, enhancing operational efficiency and decision-making. This study establishes a reproducible, resource-efficient forecasting workflow tailored specifically to the operational constraints of MSEs, filling an existing methodological gap in inventory prediction research.