基于Stacking策略的集成BN网络目标威胁评估

doi:10.12305/j.issn.1001-506X.2024.02.22

摘要/Abstract

摘要：

现有基于贝叶斯网络的威胁评估采用专家经验确定的朴素结构, 其推理评估结果精度欠佳。为此，提出一种融合专家经验与数据观测的基于Stacking策略的集成贝叶斯网络(ensemble Bayesian network, EBN)。首先使用不同搜索空间内的评分优化算法获得数据观测模型集并进行模型平均; 然后使用专家经验朴素模型对平均网络进行修剪, 形成威胁约束集合; 最后以动态规划为基础, 通过该集合限制节点序图扩展, 以求取全局最优威胁评估网络。在作战想定中, EBN模型单目标威胁概率推理精度比朴素贝叶斯模型高出10%, 在多目标威胁排序任务中, 其Spearman系数分布亦优于朴素模型。

关键词: 威胁评估, 贝叶斯网络, 结构学习, 约束优化

Abstract:

The existed threat assessment based on Bayesian networks adopts a naive structure determined by expert experience, and its inference evaluation results have poor accuracy. Thus, a Stacking strategy based ensemble Bayesian network (EBN) is proposed that integrates expert experience and data observation. Firstly, scoring optimization algorithms in different search spaces are used to obtain the data observation model set and perform model averaging. Then, expert empirical naive models are used to prune the average network to form a set of threat constraints. Finally, based on dynamic programming, the set is used to limit the expansion of the node order graph, with the aim of obtaining the global optimal threat assessment network. In combat scenarios, the EBN model has a 10% higher inference accuracy for single target threat probability than the naive Bayesian model, and the Spearman coefficient distribution is also better than the naive model in multi target threat ranking tasks.

Key words: threat assessment, Bayesian network (BN), structure learning, constrained optimization

中图分类号:

V249

王紫东, 高晓光, 刘晓寒. 基于Stacking策略的集成BN网络目标威胁评估[J]. 系统工程与电子技术, 2024, 46(2): 586-598.

Zidong WANG, Xiaoguang GAO, Xiaohan LIU. Target threat assessment based on ensemble Bayesian network with Stacking strategy[J]. Systems Engineering and Electronics, 2024, 46(2): 586-598.

图/表 20

图1

图2

图3

图4

图5

表1

表2

图6

图7

表3

表4

表5

表6

表7

表8

表9

图8

图9

图10

图11

参考文献 30

1	LIEBHABER M J, FEHER B. Air threat assessment: research, model, and display guidelines[R]. San Diego: Space and Naval Warfair Systems Command, 2002.
2	SCHUCK T , BLASCH E . OODA Loop 2.0: information not agility is life[M]. New York: Breaking Defense, 2017.
3	HINMAN M. Some computational approaches for situation assessment and impact assessment[C]//Proc. of the 5th International Conference on Information Fusion, 2022: 687-693.
4	YU D, WANG H J, YUAN H Y. Method of unknow target risk analysis and threat assessment for UUVs[C]//Proc. of the 39th Chinese Control Conference, 2020: 594-599.
5	LUO R N , HUANG S C , ZHAO Y , et al. Threat assessment method of low altitude show small (LSS) targets based on information entropy and AHP[J]. Entropy, 2021, 23 (10): 1292- 1304. doi: 10.3390/e23101292
6	PAWLAK Z . Rough sets[J]. International Journal of Computer and Information Sciences, 1982, 11 (2): 341- 356.
7	DENG Y . A threat assessment model under uncertain environment[J]. Mathematical Problems in Engineering, 2015, 2015 (15): 878024.
8	EHASN A , JAVAD H . Target threat assessment using fuzzy sets theory[J]. International Journal of Advances in Intelligent Informatics, 2015, 1 (2): 57- 74. doi: 10.26555/ijain.v1i2.18
9	BEHZADIAN M , OTAGHSARA S K , YAZDANI M , et al. A state-of the-art survey of TOPSIS applications[J]. Expert Systems with Applications, 2012, 39 (17): 13051- 13069. doi: 10.1016/j.eswa.2012.05.056
10	FAN Z P , ZHANG X , CHEN F D , et al. Extended TODIM method for hybrid multiple attribute decision making problems[J]. Knowledge-Based Systems, 2013, 43 (4): 40- 48.
11	JIANG Y P , LIANG X , LIANG H M . An I-TODIM method for multi-attribute decision making with interval numbers[J]. Soft Computing, 2017, 21 (18): 5489- 5506. doi: 10.1007/s00500-016-2139-5
12	HAUTANIEMI S K , KORPISAARI P T , SAARINEN J P P . Target identification with dynamic hybrid Bayesian networks[J]. Proc.of SPIE-The International Society for Optical Engineering, 2002, 4170, 92- 102.
13	HAUTANIEMI S K , SAARINEN J P P , DASARATHY B V . Multitarget tracking with IMM and Bayesian networks: empi-rical studies[J]. Proc.of SPIE-The International Society for Optical Engineering, 2001, 4385, 47- 57.
14	陈龙, 马亚平. 基于分层贝叶斯网络的航母编队对潜威胁评[J]. 系统仿真学报, 2017, 29 (9): 2206-2212, 2220.
	CHEN L , MA Y P . Threat assessment of aircraft carrier formation based on hierarchical Bayesian network[J]. Journal of System Simulation, 2017, 29 (9): 2206-2212, 2220.
15	高晓光, 杨宇. 基于贝叶斯网的舰艇防空威胁评估[J]. 战术导弹技术, 2020, (4): 47-57, 70.
	GAO X G , YANG Y . Threat assessment for warship air defense Based on Bayesian network[J]. Tactical Missile Technology, 2020, (4): 47-57, 70.
16	刘振, 彭军, 胡云安. 一种新型动态贝叶斯网络及其在威胁评估中的应用[J]. 火力与指挥控制, 2014, 39 (2): 16- 20.
	LIU Z , PENG J , HU Y A . A new kind dynamic Bayesian network and application to threat1 assessment[J]. Fire Control & Command Control, 2014, 39 (2): 16- 20.
17	李旭辉, 顾颖彦, 韩兴豪. 基于动态云贝叶斯网络的舰艇防空目标威胁评估[J]. 舰船电子对抗, 2021, 44 (1): 38- 44.
	LI X H , GU Y Y , HAN X H . Threat assessment of ship air defense target based on dynamic cloud Bayesian network[J]. Shipboard Electronic Countermeasure, 2021, 44 (1): 38- 44.
18	孟光磊, 龚光红. 基于混合贝叶斯网的空域目标威胁评估方法[J]. 系统工程与电子技术, 2010, 32 (11): 2398- 2401.
	MENG G L , GONG G H . Threat assessment of aerial targets based on hybrid Bayesian network[J]. Systems Engineering and Electronics, 2010, 32 (11): 2398- 2401.
19	PEARL J . Models, reasoning and inference[M]. Cambridge: Cambridge University Press, 2000.
20	LEE C, VAN B P. Metaheuristics for score-and-search Bayesian network structure learning[C]//Proc. of the 30th Canadian Conference on Artificial Intelligence, 2017: 129-141.
21	SCANAGATTA M, CORANI G, ZAFFALON M. Improved local search in Bayesian networks structure learning[C]//Proc. of the 3rd International Workshop on Advanced Methodologies for Bayesian Networks, 2017: 45-56.
22	CONSTANTINOU A C . Learning Bayesian networks with the Saiyan algorithm[J]. ACM Transactions on Knowledge Discovery from Data, 2020, 14 (4): 1- 21.
23	CONSTANTINOU A C , LIU Y , KITSON N K , et al. Effective and efficient structure learning with pruning and model averaging strategies[J]. International Journal of Approximate Reasoning, 2022, 151, 292- 321. doi: 10.1016/j.ijar.2022.09.016
24	RAMSEY J D , GLYMOUR M , SANCHEZ-ROMERO R , et al. A million variables and more: the fast greedy equivalence search algorithm for learning high-dimensional graphical causal models, with an application to functional magnetic resonance images[J]. International Journal of Data Science and Analytics, 2017, 3 (2): 121- 129. doi: 10.1007/s41060-016-0032-z
25	SAGI O , ROKACH L . Ensemble learning: a survey[J]. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 2018, 8 (5): e1249.
26	YIN J N , LI N . Ensemble learning models with a Bayesian optimization algorithm for mineral prospectivity mapping[J]. ORE Geology Reviews, 2022, 145, 104916. doi: 10.1016/j.oregeorev.2022.104916
27	COLOMBO D , MAATHUIS M H . Order-independent constraint-based causal structure learning[J]. The Journal of Machine Learning Research, 2014, 15 (1): 3741- 3782.
28	YU Y, CHEN J, GAO T, et al. Dag-GNN: dag structure learning with graph neural networks[C]//Proc. of the International Conference on Machine Learning, 2019: 7154-7163.
29	ZHENG X, DAN C, ARAGAM B, et al. Learning sparse nonparametric dags[C]//Proc. of the International Conference on Articial Intelligence and Statistics, 2020: 3414-3425.
30	CHEN Y , ZHANG T B , WANG R J , et al. A Bayesian network structural learning algorithm for calculating the failure probabilities of complex engineering systems with limited data[J]. Journal of Intelligence and Fuzzy System, 2022, 42 (3): 1991- 2004.

字段	威胁数据	字段说明
探测段	latitude	纬度
	longitude	经度
	altitude	高度/m
	angle	进入角
	speed	速度/(n mile/h)
	duration	探测持续时间/s
	detected	是否发现电磁辐射
	sources	电磁辐射源推断
	sensors	探测到目标的红方源传感器
	interval	探测到目标的时间与当前时刻间隔
	distance	目标与红方源传感器的估计距离
真实段	type	目标类型
	latitude	纬度
	longitude	经度
	altitude	高度/m
	angle	进入角
	speed	速度/(n mile/h)
	state	当前任务状态

节点	状态空间及离散化
威胁等级	3:“低威胁”; 2:“中威胁”; 1:“高威胁”; 0:“假目标”
目标类型	5:“导弹”; 4:“非隐身舰载攻击机”; 3:“电子战飞机”; 2:“无人机”; 1:“隐身舰载攻击机”; 0:“假目标”
目标状态	2:“非RTB”; 1:“RTB”; 0:“假目标”
到达时间	4:“较长”; 3:“长”; 2:“中等”; 1:“较短”; 0:“短”
电磁辐射	1:“探测到电磁辐射”; 0: “未探测到电磁辐射”
辐射源	2:“机载有源相控阵雷达”; 1:“干扰器”; 0:“未探测到辐射”
探测时间	2:“探测时间较长”; 1:“探测时间较短”; 0: “未探测到电磁辐射”
高度差	4:“较高”; 3:“高”; 2:“中”; 1:“较低”; 0:“低”
传感器	4:“其他”; 3:“有源相控阵雷达”; 2:“对空搜索雷达”; 1:“电子战装备”; 0:“电子支援/测量系统”
探测间隔	2:“长”; 1:“中”; 0:“短”
进入角	2:“大”; 1:“中”; 0:“小”
纬度差	4:“大”; 3:“较大”; 2:“中”; 1:“较小”; 0:“小”
经度差	4:“大”; 3:“较大”; 2:“中”; 1:“较小”; 0:“小”
速度	4:“快”; 3:“较快”; 2:“中”; 1:“较慢”; 0:“慢”
距离	4:“远”; 3:“较远”; 2:“中”; 1:“较近”; 0:“近”

威胁要素	方差减少量	VR/%	互信息减少量	MI/%
威胁等级	0.511 70	100	1.100 67	100
目标类型	0.451 70	88.3	0.840 98	76.4
目标状态	0.373 30	73	0.504 96	45.9
进入角	0.242 00	47.3	0.285 25	25.9
距离	0.239 20	46.7	0.450 32	40.9
速度	0.234 50	45.8	0.412 87	37.5
到达时间	0.230 80	45.1	0.457 28	41.5
经度差	0.179 60	35.1	0.295 20	26.8
纬度差	0.156 90	30.7	0.242 40	22
高度差	0.108 60	21.2	0.273 10	24.8
传感器	0.098 62	19.3	0.274 18	24.9
探测间隔	0.059 47	11.6	0.114 46	10.4
辐射源	0.050 32	9.83	0.189 58	17.2
探测时间	0.037 74	7.37	0.165 71	15.1
电磁辐射	0.036 15	7.06	0.163 73	14.9

威胁要素	方差减少量	VR/%	互信息减少量	MI/%
到达时间	2.025 0	100	2.348 11	100
速度	1.643 0	81.1	1.201 66	51.2
距离	1.471 0	72.7	0.937 75	39.9
目标类型	1.299 0	64.1	0.804 42	34.3
传感器	1.267 0	62.6	0.710 10	30.2
经度差	1.214 0	60	0.712 61	30.3
高度差	0.893 6	44.1	0.593 58	25.3
威胁等级	0.831 0	41	0.457 28	19.5
纬度差	0.799 0	39.5	0.440 98	18.8
探测时间	0.751 3	37.1	0.383 31	16.3
辐射源	0.750 7	37.1	0.380 81	16.2
电磁辐射	0.750 2	37.1	0.378 54	16.1
探测间隔	0.699 0	34.5	0.327 07	13.9
目标状态	0.393 2	19.4	0.263 62	11.2
进入角	0.253 7	12.5	0.173 17	7.37

数据集	N	PC	HC	DAG-GNN	EBN
Asia	8	-2 429.5	-2 344.8	-2 480.3	-2 344.8
Sachs	11	-8 666.2	-7 784.5	-9 188.9	-7 751.4
Child	20	-14 240.2	-12 824.2	-15 902.4	-12 772.8
Insurance	27	-16 448.9	-14 729.0	-21 415.6	-14 828.4
Alarm	37	-17 046.6	-12 777.9	-16 027.8	-11 990.6