Development of Machine Learning Strategy for Predicting the Risk Range of Ship’s Berthing Velocity

• Main Authors: Hyeong-Tak Lee, Jeong-Seok Lee, Woo-Ju Son, Ik-Soon Cho
• Format: Article
• Language: English
• Published: MDPI AG 2020-05-01
• Series: Journal of Marine Science and Engineering
• Subjects: berthing velocity; safely berth; machine learning; confusion matrix; classification
• Online Access: https://www.mdpi.com/2077-1312/8/5/376

Ships are prone to accidents when approaching in a berthing velocity greater than that allowed when determining the risk range corresponding to a port. Therefore, this study develops a machine learning strategy to predict the risk range of an unsafe berthing velocity when the ship approaches in port. To perform analysis, the input parameters were based on the factors affecting the berthing velocity, and the output parameter, i.e., the berthing velocity, was measured at a tanker terminal in the Republic of Korea. Nine machine learning classification algorithms were used to analyze each model, and the top four optimal models were selected through evaluation methods based on the confusion matrix. As a result of the analysis, extra trees, random forest, bagging, and gradient boosting classifiers were identified as good models. As a result of testing using the receiving operator characteristic curve, it was confirmed that the area under the curve of the most dangerous range of berthing velocity was the highest, thus, the risk range was appropriately classified. As such, the derived models can classify and predict the risk range of unsafe berthing velocity before approaching a port; therefore, it is possible to safely berth a ship.