Fusing XGBoost and SHAP Models for Maritime Accident Prediction and Causality Interpretability Analysis

Abstract

In order to prevent safety risks, control marine accidents and improve the overall safety of marine navigation, this study established a marine accident prediction model. The influences of management characteristics, environmental characteristics, personnel characteristics, ship characteristics, pilotage characteristics, wharf characteristics and other factors on the safety risk of maritime navigation are discussed. Based on the official data of Zhejiang Maritime Bureau, the extreme gradient boosting (XGBoost) algorithm was used to construct a maritime accident classification prediction model, and the explainable machine learning framework SHAP was used to analyze the causal factors of accident risk and the contribution of each feature to the occurrence of maritime accidents. The results show that the XGBoost algorithm can accurately predict the accident types of maritime accidents with an accuracy, precision and recall rate of 97.14%. The crew factor is an important factor affecting the safety risk of maritime navigation, whereas maintaining the equipment and facilities in good condition and improving the management level of shipping companies have positive effects on improving maritime safety. By explaining the correlation between maritime accident characteristics and maritime accidents, this study can provide scientific guidance for maritime management departments and ship companies regarding the control or management of maritime accident prevention.

A spatial-temporal analysis approach to global maritime accidents with accident point density clustering

“…Based on the results of maritime statistical analysis, with the help of a series of correlation analysis tools, the characteristics, causes and occurrence laws of accidents can be summarized [6][7] . Ma 8 established a domino effect model based on temporal association rules, analyzed the formation mechanism of accident chain, and revealed the hidden relationship behind the accident.…”

An interpretable XGBoost‐based approach for Arctic navigation risk assessment

“…It allows us to understand which features contribute most in model predictions or decisions (Liu, Chen, et al, 2022). By analyzing FI, we can identify the most significant predictors, which can provide valuable insights into underlying data and problems at hand (Zhang et al, 2022). Furthermore, the use of FI can aid in model selection and hyperparameter tuning, which enables us to develop more accurate and robust XGBoost models (Megnidio-Tchoukouegno & Adedeji, 2023).…”

“…XGBoost features efficiency (Song et al., 2020), high speed (Jin et al., 2019), outstanding performance (Jin et al., 2019), decent robustness (Jin et al., 2019), and capability of interpretations (Patidar & Tiwari, 2013). The XGBoost models have gained popularity in recent years for its ability to deliver superior performance in prediction and decision‐making (Zhang et al., 2022). It is particularly useful in assessing and predicting Arctic navigation risk due to its ability to handle large and complex data sets, and its robustness against overfitting (Rawson & Brito, 2022).…”

“…Feature importance (FI) and shapley additive explanation (SHAP) are applied to analyze the intrinsic and essential relationship between features and safety index to help us understand predictions of XGBoost models. The FI can be used to retrieve the average importance scores of all variables across all decision trees of XGBoost models (Zhang et al., 2022). It allows us to understand which features contribute most in model predictions or decisions (Liu, Chen, et al., 2022).…”

“…It enables us to gain insights into the inner workings of complex models. With SHAP, we can understand how each input feature affects the model's target and how the features interact with each other (Parsa et al., 2020; Zhang et al., 2022). In addition, SHAP provides a clear and intuitive way to communicate the model's predictions to related decision‐makers, ultimately promoting transparency and trust in the model's decision‐making process (Zhang et al., 2023).…”