Supply Chain Analytics: A Performance Evaluation of Machine Learning, Statistical, and Time-Series Models

Supply chain analytics is pivotal in enhancing supply chain performance through data driven decision-making. This study evaluates the effectiveness of various analytical models in improving supply chain performance using the DataCo Smart Supply Chain dataset. This research comprehensively evaluates machine learning, statistical, and time-series models for forecasting accuracy, demand prediction, late delivery risk, shipment duration, and route mapping optimisation. The study methodically compares and contrasts the outcomes of various modelling techniques, providing valuable insights into the most suitable approaches for optimising supply chain operations. The results indicate that decision tree and random forest models excel in supply chain forecasting and sales prediction. Similarly, Random Forest, XGBoost, and Gradient Boost models accurately predict
late delivery risk, while Exponential Smoothing, SARIMA, and ARIMA models effectively predict shipment duration. To validate these findings, rigorous statistical testing, cross-validation, and alignment with industry standards were employed, ensuring the reliability and applicability of the results. This research contributes significantly to supply chain analytics, offering practitioners and
researchers guidance on selecting appropriate methodologies for enhanced supply chain performance.