Geographical spatial analysis and risk prediction based on machine learning for maritime traffic accidents: A case study of Fujian sea area

Safety analysis according to the spatial distribution characteristics of maritime traffic accidents is critical safety management. An accident framework based on geographic information system (GIS) proposed characterize occurring in Fujian sea area 2007–2020 by employing kernel density estimation and autocorrelation techniques. The divided into various grids, each grid, mapping relationships between number severity are established. Machine learning (ML) technology used assess whether a grid an accident-prone predict grid. prediction different ML models, including random forest (RF) model, Adaboost gradient boosting decision tree (GBDT) Stacking combined were compared. optimality model was verified comparing experimental results this with those classical convolutional neural network (CNN), long short term memory (LSTM), support vector machine (SVM). According results, data set entire shows typical clustering positive correlation. That is, indicates that subareas, Ningde area, Fuzhou Xiamen generally have high densities highest risk value within whole area. High-high clustering, cluster areas neighbored other cluster, mainly seen areas, while Xiamen, Putian, Zhangzhou subareas show low-low which low clusters clusters. Among accuracy, precision, recall, F1-score values 0.912, 0.910, 0.904 predicting 0.750, 0.745, 0.746 indicating its superior performance. Based our geospatial data, method demonstrates effective reliable prediction.