Prediction of Grades of Ship Collision Accidents Based on Random Forests and Bayesian Networks

Abstract

Ship collision accidents are typical and major ones for ships, whose grades are predicted to be favorable for taking timely measures and relieving the corresponding losses or reducing their occurrence possibilities. To this end, a model based on Random Forests and Bayesian Network Model was put forward here to predict the grade of any ship collision accident; the former were utilized to identify key factors influencing prediction of ship collision accident grades while the identified results acted as nodes of the latter. By taking 945 ship collision accidents in Jiangsu Section in the Main Stem of Yangtze River, the Bayesian network model was constructed by means of machine learning to predict the collision grades.