Improving shipping container damage claims prediction through level 4 information fusion

Abstract

Maritime trade accounts for approximately 90% of global trade, and most global supply chains incorporate some form of maritime travel (Cheraghchi et al., 2017). Optimising operations at commercial maritime ports is therefore of significant importance worldwide, and impacts global trade. While damage to vessels and cargo has been studied extensively, as has optimising portside operational efficiency, investigations of damage to shipping containers themselves and the resultant disruption of port-side efficiency remains unstudied. The application of machine learning (ML) techniques to uncover causes of shipping container damage allows for more efficient handling of the service-disruptions they cause, as well as insights into the veracity of the current wisdom held by domain experts in the industry. Further, the application of ML methodologies for dynamic algorithm selection (per level 4 of the JDL/DFIG data fusion model) allows for significant improvements to the overall performance of the aforementioned ML methodologies.