Data for Backorder time prediction in a DDMRP managed production system

In the interest of transparency and reproducibility, the full dataset used for backorder time prediction in a DDMRP production model, along with the associated machine learning model code, has been made publicly available. The resulting dataset, generated from the simulation of all scenarios, comprises 10,620 rows and 15 columns. The target variable is the backorder time of a delayed customer order in a DDMRP-managed system, measured from the moment the order is received until it is fulfilled. The input features were selected based on insights from the literature and include key factors known to influence backorder times in DDMRP systems. Instructions: Preprocess the dataset and split it into training, validation, and testing subsets. Use grid search with K-fold cross-validation on the training data to identify optimal hyperparameters and reduce overfitting. Train multiple machine learning models (e.g., Linear Regression, DT, RF, XGBoost, SVM, and ANNs) using the tuned hyperparameters. Compare model performance to determine the most accurate approach for backorder time prediction in a DDMRP-managed flowshop.