Data-driven models for optimizing second-hand ship trading strategies under contextual information

Abstract

Second-hand ship online trading platforms (SOTPs) are reshaping the traditional broker-reliant second-hand ship transactions. This study investigates the decision-making process within the context of SOTP from a shipowner's perspective. We introduce a comprehensive framework tailored to guide shipowners in strategically navigating pivotal decisions, including the adoption of SOTP and the specification of optimal minimum starting prices while leveraging the value of online transaction data. Our approach is rooted in data-driven decision-making under uncertainty, employing quantile regression forests (QRF), and weighted sample average approximation (wSAA). The latter encompasses a predictive wSAA model, a local wSAA model, and a residual-based wSAA model. Each of these models provides a unique perspective on weight determination within the wSAA paradigm. To validate our proposed approach, we draw upon extensive real-world data sourced from a Chinese SOTP between January 2017 and May 2023. Within this context, our numerical experiments pursue three primary objectives: (i) identifying performance disparities among the models, (ii) assessing the value of contextual information, and (iii) evaluating the optimal strategy for shipowners. Our findings not only underscore the efficacy of our approaches but also provide invaluable insights into the adoption of SOTPs, establishing a robust foundation for informed decision-making in the continually evolving SOTP landscape.