선박 접안속도의 K-평균 군집분석을 활용한 도선사 조선 유형에 관한 연구

Abstract

The most important factor when a ship is berthing is the berthing velocity. Excessive berthing velocity may cause accidents in ports and ships, so it is necessary to be cautious. Since the ship’s berthing is completed by the pilot on board, the propensity of the pilots is closely related to the decision of the berthing velocity. Therefore, this study the pattern of pilot's maneuvering to berth the ship to prevent berthing accidents was analyzed by clustering. K-means algorithm was used as a clustering analysis method as one of the unsupervised learning of machine learning. The berthing velocity data were collected for the pilot cluster analysis, the data were collected for 39 months at a domestic tanker jetty. During the collection period, 47 pilots were surveyed who boarded the ship that berthed at the jetty. The 47 pilots were grouped by individual and analyzed basic data including the average and standard deviation of the berthing velocity. Based on the analyzed mean and standard deviation, the k-means algorithm was applied for the clustering of pilots. As a result, the pilots were clustered into four groups. Groups with low mean and standard deviation, medium mean but low standard deviation, groups with medium mean and standard deviation, and finally, group with relatively high mean and standard deviation. Based on the four groups, the pilots were divided into three patterns of maneuvering: Low, Moderate, and High Risk.