A data-driven optimization approach to improving maritime transport efficiency

Ship inspections conducted by port state control (PSC) can effectively reduce maritime risks and protect the marine environment. The effectiveness of PSC depends on accurately selecting ships with higher risk for inspection. Ship risk profile (SRP) is currently the most common method of quantifying ship risk, but the thresholds of the factors that determine a ship's risk and classification in the SRP framework are subjective and can make the ship selection process less efficient. In this study we propose a data-driven bi-objective nonlinear programming model, referred to as the SRP+ model, to optimize the thresholds in the original SRP framework. To solve the model, we first linearize the nonlinear optimization model using the big-M method, and then develop an augmented epsilon-constraint method to transform the bi-objective model to a single-objective model and obtain all Pareto optimal solutions. We also conduct a case study using real PSC inspection records at the Hong Kong port to construct and validate the SRP+ model. The results suggest that the threshold of the total weighting points to classify a ship as high-risk ship should be slightly increased, the thresholds of ship age should be significantly increased, the threshold of historical deficiency number should be increased, while the threshold of historical ship detention times should be decreased. The proposed SRP+ model can benefit both conservative and open-minded port authority decision makers by identifying ships with more deficiencies and/or higher detention probability more efficiently. The model can also be applied to other risk management problems in transportation and supply chain management, in addition to the maritime transport domain.