Prediction of task occurrence distribution for automated shop floor planning using multi-output support vector regressor

Abstract

The digitalization of shop floors has led to a significant shift towards automated planning and scheduling to improve resource management and production efficiency. This paper presents a comparative study of the use of machine learning approaches for predicting the distribution of task occurrence in activity-based shop floors. This study leverages real data extracted from Enterprise Resource Planning (ERP) and Manufacturing Execution Systems (MES), and historical data of shopfloor-level processes. Furthermore, three regression-based models, namely a fe-nearest Neighbor Regressor (KNR), Random Forest Regressor (RFR), and Multi-output Support Vector Regressor (M-SVR) are evaluated on the extracted data. The study identifies M-SVR as the best-performing model when hyperparameters were optimised through model optimisation via grid search and 5-fold cross-validation. The comparative analysis includes evaluation metrics, providing insight into effective task prediction in shop floor environments. This paper highlights the importance of data-driven methods for the prediction of manufacturing processes and the digitalization of shop floors.