基于树增强朴素贝叶斯网络的关键海域事故风险评价模型

doi:10.16265/j.cnki.issn1003-3033.2025.12.0529

摘要/Abstract

摘要：

为应对关键海域事故频发且危害较大的问题,构建一种基于树增强朴素贝叶斯(TAN)网络的关键海域事故风险评价模型;为克服事故报告中存在部分样本偏差的缺陷,利用箱线图方法消除异常值,以提升数据质量;考虑风险因素复杂性以及相关性特征,利用随机森林算法,识别出关键风险因素,并构建关键海域事故风险评价指标体系;此外,将TAN网络模型与6种机器学习模型对比,进行验证分析。结果表明:TAN网络模型的准确率最优,为93.02%;船速、船舶长度、海盗袭击是诱发关键海域风险事件的主要因素;船龄在11~20年的船舶应作为重点维修与检验对象;在水深较浅的关键海域航行的船舶应加强操作谨慎性。

关键词: 树增强朴素贝叶斯(TAN)网络, 关键海域, 海上事故, 风险评价, 随机森林

Abstract:

In response to the frequent and high-impact accidents in key marine areas, a risk assessment model for accidents in such areas based on TAN network was established. To address the issue of partial sample bias in accident reporting, the boxplot method was employed to eliminate outliers and improve data quality. Considering the complexity and correlation of risk factors, a random forest algorithm was utilized to identify key risk factors and establish a risk evaluation index system for accidents in key marine areas. In addition, the performance of TAN network model was compared with six machine learning models for validation and analysis. The results demonstrate that TAN network achieves the highest accuracy of 93.02%. The findings indicate that ship speed, ship length, and pirate attacks are the primary factors contributing to risk events in key marine areas. Vessels aged between 11 and 20 years should be prioritized for maintenance and inspection. In addition, ships navigating in shallow key marine areas should operate with increased caution.

Key words: tree augmented naive Bayesian (TAN) network, key marine areas, maritime accidents, risk assessment, random forest

中图分类号:

X913.4安全系统工程

吕靖, 任梓昕, 范瀚文, 常征. 基于树增强朴素贝叶斯网络的关键海域事故风险评价模型[J]. 中国安全科学学报, 2025, 35(12): 172-179.

LYU Jing, REN Zixin, FAN Hanwen, CHANG Zheng. Risk evaluation model of accidents in key marine areas using tree augmented naive Bayesian network[J]. China Safety Science Journal, 2025, 35(12): 172-179.

图/表 9

图1

图2

图3

图4

图5

表1

表2

表3

表4

参考文献 16

[1]	CAO Xinhu, LAM L S J. A fast reaction-based port vulnerability assessment: case of Tianjin port explosion[J]. Transportation Research Part A, 2019, 128: 11-33.
[2]	GU Bingmei, LIU Jiaguo. Port resilience analysis based on the HHM-FCM approach under COVID-19[J]. Ocean and Coastal Management, 2023, 243: DOI: 10.1016/j.ocecoaman.2023.106741.
[3]	陈伟炯, 苏子薇, 秦庭荣, 等. 采用云重心评判法的港口船舶溢油综合应急能力评价[J]. 中国安全科学学报, 2024, 34(4): 183-190. doi: 10.16265/j.cnki.issn1003-3033.2024.04.1765
	CHEN Weijiong, SU Ziwei, QIN Tingrong, et al. Evaluation of comprehensive emergency response capability for port ship oil spill with cloud gravity center evaluation method[J]. China Safety Science Journal, 2024, 34(4): 183-190. doi: 10.16265/j.cnki.issn1003-3033.2024.04.1765
[4]	GERGES A M. The Red Sea and Gulf of Aden action plan:facing the challenges of an ocean gateway[J]. Ocean and Coastal Management, 2002, 45(11): 885-903. doi: 10.1016/S0964-5691(02)00112-6
[5]	李邓宇卉, 焦建彬, 汪寿阳, 等. 全球关键航运通道受阻对供应链的影响:“长赐号”在苏伊士运河搁浅案例分析[J]. 管理评论, 2025, 37(2): 237-248.
	LI Dengyuhui, JIAO Jianbin, WANG Shouyang, et al. The impact of global key shipping channel disruption on supply chain:case study on the grounding of ever given in Suez Canal[J]. Management Review, 2025, 37(2): 237-248.
[6]	FAN Shiqi, YANG Zaili, WANG Jin, et al. Shipping accident analysis in restricted waters: lesson from the Suez Canal blockage in 2021[J]. Ocean Engineering, 2022, 266: DOI: 10.1016/j.oceaneng.2022.113119.
[7]	范瀚文, 常征, 王聪. 海上运输通道关键节点安全韧性影响因素及评价[J]. 上海海事大学学报, 2022, 43(2):40-46.
	FAN Hanwen, CHANG Zheng, WANG Cong. Influencing factors and evaluation of safety resilience of key nodes in maritime transportation channels[J]. Journal of Shanghai Maritime University, 2022, 43(2): 40-46.
[8]	李宝德, 吕靖, 李晶. 基于改进Copula的海上运输关键节点应急保障能力评估[J]. 交通运输系统工程与信息, 2020, 20(2): 20-25.
	LI Baode, LYU Jing, LI Jing. Evaluation of emergency guarantee capability of key nodes in maritime transport based on improved Copula function[J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(2): 20-25.
[9]	BÜNYAMIN K, ERKAN Ç. Data-driven Bayes approach on marine accidents occurring in Istanbul strait[J]. Applied Ocean Research, 2022, 123: DOI: 10.1016/j.apor.2022.103180.
[10]	陈伟炯. 船舶航行安全的可持续改善途径[J]. 交通运输工程学报, 2002, 2(3): 115-119.
	CHEN Weijiong. Methods for sustainably improving ships navigation safety[J]. Journal of Traffic and Transportation Engineering, 2002, 2(3): 115-119.
[11]	范瀚文, 吕靖. 基于规则推理贝叶斯网络的海上通道关键节点韧性评价[J]. 安全与环境学报, 2024, 24(11): 4121-4129.
	FAN Hanwen, LYU Jing. Resilience assessment of key nodes along sea lanes of communication using rule-based Bayesian networks[J]. Journal of Safety and Environment, 2024, 24(11): 4121-4129.
[12]	李苗, 李凌波, 左志恒, 等. 基于机器学习的成品油管道运行工况识别[J]. 中国安全科学学报, 2024, 34(6): 127-135. doi: 10.16265/j.cnki.issn1003-3033.2024.06.1410
	LI Miao, LI Lingbo, ZUO Zhiheng, et al. Machine learning-based recognition for recognizing operating conditions of multi-product pipelines[J]. China Safety Science Journal, 2024, 34(6): 127-135. doi: 10.16265/j.cnki.issn1003-3033.2024.06.1410
[13]	李雨阳, 刘舫, 胡文君, 等. 基于多源遥感数据的植被精细分类方法对比[J]. 世界地质, 2025, 44(2): 311-321.
	LI Yuyang, LIU Fang, HU Wenjun, et al. Comparation on vegetation fine classification methods based on multi-source remote sensing data[J]. World Geology, 2025, 44(2): 311-321.
[14]	阎莹, 王玉莹, 周旋, 等. 基于混合模型的自动驾驶车辆事故严重程度影响因素分析[J]. 交通运输工程学报, 2025, 25(1): 184-196.
	YAN Ying, WANG Yuying, ZHOU Xuan, et al. Analysis of autonomous vehicle accident severity factors based on hybrid model[J]. Journal of Traffic and Transportation Engineering, 2025, 25(1): 184-196.
[15]	JIANG Meizhi, LU Jing. Maritime accident risk estimation for sea lanes based on a dynamic Bayesian network[J]. Maritime Policy & Management, 2020, 47(5): 649-664.
[16]	LI Huanhuan, ZHOU Kaiwen, ZHANG Chao, et al. Dynamic evolution of maritime accidents: comparative analysis through data-driven Bayesian networks[J]. Ocean Engineering, 2024, 303: DOI: 10.1016/j.oceaneng.2024.117736.

实际结果	预测结果
实际结果	非常严重	严重	不太严重
非常严重	38	2	0
严重	4	34	0
不太严重	0	0	8

性能指标	非常严重	严重	不太严重
准确率	0.930	0.930	1.000
精确率	0.950	0.895	1.000
召回率	0.905	0.944	1.000
F₁	0.927	0.919	1.000
特异性	0.955	0.920	1.000
假正例率	0.045	0.080	0.000

模型	准确率/ %	变化/ %	Kappa 系数	变化
TAN网络	93.02	—	0.879	—
极端随机树	83.72	9.30	0.712	0.167
梯度提升树	76.74	16.28	0.580	0.299
K最近邻	74.42	18.60	0.541	0.338
朴素贝叶斯	61.63	31.40	0.311	0.568
支持向量机	56.98	36.05	0.188	0.691
反向传播神经网络	54.65	38.37	0.184	0.695

评价指标	真实风险影响			平均值	排名
评价指标	非常严重	严重	不太严重	平均值	排名
S₂	15.35	18.90	11.03	15.09	1
S₃	11.40	14.95	11.40	12.58	2
N₁	18.55	15.15	3.48	12.39	3
S₅	14.60	8.35	12.81	11.92	4
E₂	16.05	14.25	1.85	10.72	5
S₁	12.25	12.15	4.16	9.52	6
G₁	10.75	12.95	3.65	9.12	7
E₇	13.35	10.25	3.08	8.89	8
G₄	10.35	11.80	1.99	8.05	9
S₇	8.00	10.30	3.96	7.42	10
E₆	7.70	9.00	4.29	7.00	11
S₄	4.05	8.95	6.73	6.58	12
E₁	6.55	9.75	3.18	6.49	13
E₃	9.05	8.35	0.70	6.03	14
G₃	4.15	7.40	5.37	5.64	15
G₂	8.05	5.45	2.60	5.37	16
G₆	5.90	7.25	1.35	4.83	17
E₈	3.85	4.50	0.55	2.97	18