Enhancing Supply Chain Efficiency Through Machine Learning: A Predictive Analytics Approach to Risk Identification and Timely Deliveries

Abstract

The present study covers risk identification, development of on-time delivery with the help of machine learning, and predictive analytics for further improvements in supply chain management. Supply chains across the world have turned complex due to the number of involved stakeholders. The after-effect of this factor has been that even minor interruptions of delivery have produced poor results and unexpected dangers that can cause serious harm to operational efficiency and customer satisfaction. In this study, Random Forests, Logistic Regression, and Neural Networks were used. These models enable the use of machine-learning algorithms, which are able to predict possible risks, including those of delayed delivery, using past and current data from supply chain systems. The dataset used in the study was divided into 144,415 training samples and 36,104 test samples, where each sample consists of six features corresponding to six significant variables of the supply chain. The metrics used for the evaluation of the models are precision, recall, accuracy, and AUC-ROC. The Random Forest model developed an accuracy of 97.04% and an AUC score of 0.9728, which shows that it is highly predictive based on those metrics. This performance indicates the model's efficiency in predicting late delivery, which would mean giving supply chain managers critical information they need to drive more-informed decisions toward real-time proactiveness in risk reduction. This paper focuses on the practical use of machine learning in supply chain risk management and proposes a framework integrated with real-time data to optimize operational efficiency for on-time delivery.