Prediction of the severity of marine accidents using improved machine learning

Transportation Research Part E Logistics and Transportation Review, Год журнала: 2024, Номер 188, С. 103647 - 103647

Опубликована: Июль 2, 2024

Although many studies have focused on the occurrence likelihood of marine accidents, few analysis severity consequences, and even fewer prediction severity. To this end, a new research framework is proposed in study to accurately predict accidents. First, novel two-stage feature selection (FS) method was developed select rank Risk Influential Factors (RIFs) improve accuracy Machine Learning (ML) model interpretability FS. Second, comprehensive evaluation measure performance FS methods based stability, predictive improvement, statistical tests. Third, six well-established ML models were used compared different predictors. The Light Gradient Boosting (LightGBM) found best for accidents treated as benchmark model. Finally, LightGBM accident RIFs selected by method, effect risk control measures counterfactually analysed from quantitative perspective. This innovative use improved approaches can effectively analyse providing methodology triggering direction using Artificial Intelligence (AI) technologies safety assessment prevention studies. source code publicly available at: https://github.com/FengYinLeo/PGI-SDMI.

Язык: Английский

---

Dynamic evolution of maritime accidents: Comparative analysis through data-driven Bayesian Networks

Ocean Engineering, Год журнала: 2024, Номер 303, С. 117736 - 117736

Опубликована: Апрель 10, 2024

Maritime accident research has primarily focused on characteristics and risk analysis, which often overlooks the evolution of associated patterns over time. This study aims to investigate dynamic changes in maritime accidents from 2012 2021 by employing a data-driven Bayesian Network (BN) model conducting systematic pattern comparison. It presents two-stage models for two databases five against different timeframes capture evolving global accidents. Furthermore, within context investigation, this pioneers analysis effectiveness network structures, namely layered BN Tree-Augmented Naive (TAN) network, terms accuracy predicting severity. The key findings regarding past decade include: (1) significant rise risks linked large ships (30.8%), port areas (11.67%), anchoring (11.82%), manoeuvering operations (3.8%); (2) connection between poor practices fishing boats 'overboard' accidents, inadequate equipment tankers or chemical 'fire/explosion' accidents; (3) TAN model's superior performance forecasting severity compared model; (4) probability 'very serious' ship-related factors is 74.7%, significantly lower than network's 99.4%. reveals shifts time underscores importance continuous monitoring effective safety management.

Язык: Английский

---

Incorporation of a global perspective into data-driven analysis of maritime collision accident risk

Reliability Engineering & System Safety, Год журнала: 2024, Номер 249, С. 110187 - 110187

Опубликована: Май 15, 2024

Ship collision accidents are one of the most frequent accident types in global maritime transportation. Nevertheless, conducting an in-depth analysis prevention poses a formidable challenge due to constraints limited Risk Influential Factors (RIFs) and available datasets. This paper aims incorporate perspective into new data-driven risk model, scrutinize root causes accidents, advance measures for their mitigation. Additionally, it seeks analyze spatial distribution conduct comprehensive comparative study on characteristics both pre- post-COVID-19, utilizing real dataset collected from two reputable organizations: Global Integrated Shipping Information System (GISIS) Lloyd's Register Fairplay (LRF). The research findings implications encompass several crucial aspects: 1) constructed model demonstrates its reliability accuracy predicting as evident prediction performance various scenario analysis; 2) hazardous voyage segment is identified provide valuable guidance different stakeholders; 3) hierarchical significance ship context highlighted regarding probable occurrences; 4) During pandemic, rise probabilities, particularly involving older vessels bulk carriers, implies heightened operational challenges or maintenance issues these types; (5) prominence favorable adverse sea conditions reports underscores significant influence weather during pandemic. These help enhance safety protocols, ultimately reducing frequency domain.

Язык: Английский

---

Risk analysis of lithium-ion battery accidents based on physics-informed data-driven Bayesian networks

Reliability Engineering & System Safety, Год журнала: 2024, Номер 251, С. 110294 - 110294

Опубликована: Июнь 24, 2024

Язык: Английский

---

Navigating uncertainty: A dynamic Bayesian network-based risk assessment framework for maritime trade routes

Reliability Engineering & System Safety, Год журнала: 2024, Номер 250, С. 110311 - 110311

Опубликована: Июль 1, 2024

Язык: Английский

---

Investigation of ship collision accident risk factors using BP-DEMATEL method based on HFACS-SCA

Reliability Engineering & System Safety, Год журнала: 2025, Номер 257, С. 110875 - 110875

Опубликована: Янв. 31, 2025

Язык: Английский

---

A knowledge graph-based hazard prediction approach for preventing railway operational accidents

Reliability Engineering & System Safety, Год журнала: 2024, Номер 247, С. 110126 - 110126

Опубликована: Апрель 4, 2024

Язык: Английский

---

Analyzing risk influencing factors of ship collision accidents: A data-driven Bayesian network model integrating physical knowledge

Ocean & Coastal Management, Год журнала: 2024, Номер 256, С. 107311 - 107311

Опубликована: Июль 30, 2024

Язык: Английский

---

Maritime near-miss prediction framework and model interpretation analysis method based on Transformer neural network model with multi-task classification variables

Reliability Engineering & System Safety, Год журнала: 2025, Номер unknown, С. 110845 - 110845

Опубликована: Янв. 1, 2025

Язык: Английский

---

Decoding dependencies among the risk factors influencing maritime cybersecurity: Lessons learned from historical incidents in the past two decades

Ocean Engineering, Год журнала: 2024, Номер 312, С. 119078 - 119078

Опубликована: Авг. 29, 2024

The distinctive features of maritime infrastructures present significant challenges in terms security.Disruptions to the normal functioning any part transportation can have wide-ranging consequences at both national and international levels, making it an attractive target for malicious attacks.Within this context, integration digitalization technological advancements seaports, vessels other elements exposes them cyber threats.In response critical challenge, paper aims formulate a novel cybersecurity risk analysis method ensuring security.This approach is based on data-driven Bayesian network, utilizing recorded incidents spanning past two decades.The findings contribute identification highly contributing factors, meticulous examination their nature, revelation interdependencies, estimation probabilities occurrence.Rigorous validation developed model ensures its robustness diagnostic prognostic purposes.The implications drawn from study offer valuable insights stakeholders governmental bodies, enhancing understanding how address threats affecting industry.This knowledge aids implementation necessary preventive measures.

Язык: Английский

---