A decision-oriented cost-sensitive learning approach for optimal decisions in ship management

This study focuses on shipowners’ decisions within the second-hand ship online trading platform (SOTP), recognizing that second-hand ship trading is a crucial entry point into the maritime industry. We investigate two key decisions of shipowners’ optimal choices: whether to sell vessels on the platform and how to set the minimum bidding increment. Leveraging the problem structure, we propose a two-stage end-to-end learning approach based on machine learning models. The proposed framework explicitly incorporates economic decision loss, addresses cost-sensitive decision-making, and improves aggregate consistency between predicted and observed outcomes through a decision-oriented calibration mechanism. By collecting real-world data, we conduct a series of experiments to evaluate the effectiveness of the proposed approach. The results demonstrate that the approach is both effective and robust across a range of settings, achieving more than 90% reduction in decision loss relative to benchmark methods, thereby significantly enhancing secondary market efficiency. In addition, we empirically analyze decision rules for setting the minimum bidding increment. Given the low sunk costs associated with vessel sales, we recommend that shipowners actively participate in trading on the SOTP. Our research provides a data-driven framework for understanding the dynamics of fleet renewal and guides intelligent decision-making within the maritime industry.