Data analytics in transport: Does Simpson's paradox exist in rule of ship selection for port state control?

Abstract

<abstract> <p>Although previous studies have applied artificial intelligence techniques to improve the accuracy and efficiency of ship selection in port state control (PSC) inspections, the new inspection regime (NIR) is still in effect and widely adopted by PSC authorities in the Tokyo Memorandum of Understanding to select ships for inspection. It considers seven features, and each candidate value of a certain feature is assigned a fixed weighting point. The sum of the weighting points of these seven features determines the risk level of a ship. The assumption behind the NIR is that ships with values attached with higher weighting points should have more deficiencies. However, this paper finds that Simpson's paradox may exist for this assumption; that is, the average number of deficiencies of ships with values attached with higher weighting points is lower than that of ships with values attached with lower weighting points. Therefore, this paper examines the plausibility of the NIR's weighted-sum method and further explores which feature flips the effect. Finally, we arrive at the conclusion that the features selected by NIR are coupled with each other, so we should not use a simple weighted-sum method to determine the risk level of a candidate ship. Based on the results, we further provide suggestions for PSC authorities with respect to the improvement of the ship selection scheme of NIR.</p> </abstract>