Research on the Risk of Water Traffic Accidents Considering Risk Navigation Preference Behavior

The prediction of ship traffic accidents has always been a hot topic in the industry. Although the traffic safety risk assessment model based on waterway operation status has achieved certain results, it is still unsatisfactory. Therefore, two traffic accident prediction models have been built in the article based on logistic regression algorithm. The first model only considers the main characteristics of waterway operation status as variables, while the second model considers both risk navigation preference behavior and the main characteristics of waterway operation status as variables. Actual data verification shows that compared with the prediction results of the first model, accuracy of the second model has increased by 5.20%, false alarm rate has decreased by 2.51%, false alarm rate has decreased by 6.62%, and Area Under Curve (AUC) value has increased by 0.04. Research results indicate that using the main characteristics of risk navigation preference behavior as variables in traffic accident prediction models yields better results.