A hybrid ensemble learning-based prediction model to minimize delay in air cargo transport using bagging and stacking

Abstract

Manufacturing productivity is inextricably linked to air freight handling for the global delivery of finished and semi-finished goods. In this article, our focus is to capture the transport risk associated with air freight which is the difference between the actual and the planned time of arrival of a shipment. To mitigate the time-related uncertainties, it is essential to predict the delays with adequate precision. Initially data from a case study in the transportation and logistics sector were pre-processed and divided into categories based on the duration of the delays in various legs. Existing datasets are transformed into a series of features, followed by extracting important features using a decision tree-based algorithm. To predict the delay with maximum accuracy, we used an improved hybrid ensemble learning-based prediction model with bagging and stacking enabled by characteristics like time, flight schedule, and transport legs. We also calculated the dependency of accuracy on the point in time during business process execution is examined while predicting. Our results show all predictive methods consistently have a precision of at least 70 percent, provided a lead-time of half the duration of the process. Consistently, the proposed model provides strategic and sustainable insights to decision-makers for cargo handling.