Delay Prediction to Mitigate E-commerce Supplier Disruptions using Voting Mechanism

Abstract

COVID-19 has severely affected supply chains and last-mile logistics. With a remarkable expansion in the internet-based deals of B2C online business and changed customer behaviors, it is critical to estimate delays with sufficient precision to avoid time-related uncertainty. The risk of a supplier's delivery being late will harm businesses' ability to fulfill client orders. Shipment information of an online business organization is utilized for the investigation. The dataset is pre-processed and important features are extracted using the Random Forest technique. We propose an enhanced hybrid voting-based classification model with Trees, and Ensemble techniques (like bagging and boosting) enabled by parameters like shipping mode, scheduled shipment time, and order type to anticipate the postponement with the highest accuracy. Since the base classifiers in the voting mechanism cannot perform at the same level, we assigned various weights and noticed a significant improvement in classification performance. Consistently, the proposed model depicts improved performance and provides strategic, operational, and industrial insights for decision-making in last-mile businesses.