Modeling human's collision avoidance direction in an encounter situation using an ensemble classifier

Predicting and responding to a vessel's collision avoidance course is crucial for navigational safety. To predict the course of deciding collision avoidance, the results of human decisions should be modeled to avoid collision in a potentially collision-prone situation. This study derives a supervised learning-based ensemble classification model to model humans' decisions to avoid collision. To illustrate the model, ship trajectory data were collected and used to construct predictor and outcome variables. Figures obtained from the collision-avoidance algorithm were included in the prediction variables. These variables were applied to the classification model. Thus, a well-fitted model with high performance was constructed. This model provides insight to predict the relationship between an avoidance direction output by a collision avoidance algorithm and the variables used in the algorithm. The results indicate that the proposed model can be used to predict the direction of avoidance in a given situation.