IMPROVING DEMAND FORECASTING ACCURACY FOR PRODUCTION CAPACITY PLANNING

Abstract

This study examines approaches to improve demand forecasting accuracy in production capacity planning through a comparative analysis of statistical, machine learning, and deep learning models. Accurate forecasting is crucial for optimizing production schedules, reducing inventory costs, and ensuring efficient resource utilization. However, traditional methods often struggle with
nonlinear demand patterns and market volatility, leading to higher prediction errors. In this research, classical models such as ARIMA and Holt-Winters are compared with machine learning models including Random Forest, Support Vector Regression,
and Gradient Boosting, as well as a Long Short-Term Memory (LSTM) deep learning model. The results show that machine learning and deep learning methods significantly outperform traditional statistical approaches in MAE, RMSE, and MAPE metrics. Among all models, LSTM achieves the highest accuracy due to its ability to capture long-term dependencies in time-series data. The findings confirm that advanced data-driven methods enhance forecasting reliability and support better production planning decisions.