Machine Learning-Enabled Production Planning for Intelligent SAP Manufacturing Systems

Production planning is a critical component of intelligent manufacturing, where accurate forecasting and efficient resource allocation directly influence operational performance, inventory utilisation, and enterprise sustainability. Modern SAP Enterprise Resource Planning (ERP) systems provide integrated planning capabilities; however, conventional planning methods remain largely dependent on rule-based logic and historical data, limiting their ability to respond to dynamic manufacturing environments. This paper proposes Machine Learning-Enabled Production Planning for Intelligent SAP Manufacturing Systems, an intelligent decision-support framework designed to enhance production planning using advanced predictive analytics. The proposed framework is evaluated using the Multi-Stage Continuous-Flow Manufacturing Process dataset comprising 14,088 production records and 116 operational features. The methodol-ogy incorporates comprehensive data preprocessing, including missing-value treatment and YeoJohnson power transformation, followed by SAP-oriented feature engineering that derives production planning indicators, statistical descriptors, rolling-window variables, lag-based features, and manufacturing performance metrics. Feature selection using SelectKBest with Mutual Information Regression identifies the 100 most informative predictors for model development. Two ensemble learning algorithms, Random Forest and XGBoost, are optimised and evaluated for production performance prediction. Experimental results demonstrate that the optimised Random Forest model achieves superior predictive performance with R2 = 85.91%, RMSE = 1.4033, MAE = 1.1055, and MAPE = 1.53%, outperforming XGBoost in both prediction accuracy and model stability. To improve transparency and support enterprise decision-making, the framework incorporates explainable artificial intelligence techniques using SHAP and LIME for global and local model interpretation, while scenario-based evaluation and Monte Carlo simulation are employed to assess robustness under uncertainty. The findings demonstrate that integrating machine learning with SAP production planning enables more intelligent, scalable, interpretable, and data-driven manufacturing systems, supporting improved production efficiency, optimised inventory management, and sustainable enterprise operations.Keywords: SAP, demand planning, predictive analytics, inventory management, enterprise resource planning, machine learning.