South-Central Minzu University,a. College of Electronic and Information Engineering; b. Hubei Key Laboratory of Intelligent Wireless Communications,Wuhan 430074,China
The traditional risk assessment methods relying on expert knowledge are usually static, which cannot meet the needs of identification and assessment of variable risks of immovable cultural heritage. A dynamic risk assessment method for cultural heritage security based on semantic path coding heterogeneous graph network is proposed. By constructing a cultural heritage security knowledge graph, the influencing factors related to the dynamic risk of cultural heritage security are stored in a normalized way, and then the entity features in the security knowledge graph are aggregated using semantic path coding, which helps the network to extract a more accurate representation of the entity knowledge, and improves the accuracy of the risk level classification. The method is tested on several national key cultural heritage protection units, and the accuracy rate of risk level classification is 94.74%, which is basically consistent with the results of expert assessment, and it has certain guidance and reference value for risk control and security management of cultural heritage.
定义1:文物安防知识图谱通过 G = ( E,)来表示,其中 E = {E1,E2,…,Ei,…,EN }代表了对应知识图谱内的实体集合,i表示实体Ei 的索引,i∈[1,N],N代表文物安防知识图谱 G 中的实体总数; R = {R1,R2,…,Ri,…,RM }代表了对应知识图谱内的关系集合,M代表文物安防知识图谱 G 中的关系总数.
定义2:在文物安防知识图谱中,存在着许多种类的实体以及它们所相应的不同实体属性特征.实体类型可以通过 = {=文物保护单位,=防护对象,=人力防范,=实体防范,=技术防范}来表示,实体属性通过 H = {,,…,,…,}∈ RN×Fi 来表示,和Fi 分别代表了特定的实体Ei 的属性特征向量和特征向量对应的特征维度.
定义3:在文物安防知识图谱中,同样存在着许多不同类型的关系,在每种关系下,实体所包含的语义信息是不一样的,这些不同的语义关系可以通过 = {1,2,…,n,…,P }来表示,n∈[1,P],P代表文物安防知识图谱 G 中的关系类型的数量,n代表了第n种特定的语义关系,n代表了第n特定的语义关系下的一个实例.实体类型与通过语义关系n连接起来,构成形如的路径就被称为语义路径.如图6所示,例如防护对象-技术防范(OT),代表技术防范实体和防护对象实体之间设置关系构成的语义路径.
将文物安防知识图谱的实体特征矩阵和表示实体间连接关系的邻接矩阵作为文物安防系统风险评估模型的输入.模型结构共分为三层:基于语义路径编码的实体级注意力层、语义级注意力层和实体分类层.在文物安防知识图谱 G = ( E, R )中,实体间具有不同的语义关系,给定一种语义关系,每个实体都有基于该语义关系的邻居实体.首先,由于图谱中实体的异构性,使得各类实体所对应的实体属性特征空间也不尽相同.因此需要先将各类实体的属性特征投影到相同的特征空间内;通过对实体属性特征作线性变换,以获取实体属性特征在高维空间里的嵌入表示:
PALTRINIERIN, COMFORTL, RENIERSG. Learning about risk: Machine learning for risk assessment[J]. Safety Science, 2019, 118: 475-486.
[7]
BONDEA, BONDEM, TROELSENA, et al. Assessing the utility of a sliding-windows deep neural network approach for risk prediction of trauma patients[J]. Scientific Reports, 2023, 13(1): 5176.
JIS X, PANS R, CAMBRIAE, et al. A survey on knowledge graphs: Representation, acquisition, and applications[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(2): 494-514.
[10]
ZHANGZ W, CUIP, ZHUW W. Deep learning on graphs: A survey[J]. IEEE Transactions on Knowledge and Data Engineering, 2022, 34(1): 249-270.
[11]
LIZ F, LIUH, ZHANGZ L, et al. Learning knowledge graph embedding with heterogeneous relation attention networks[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(8): 3961-3973.
[12]
LIUZ G, QIANP, WANGX Y, et al. Combining graph neural networks with expert knowledge for smart contract vulnerability detection[J]. IEEE Transactions on Knowledge and Data Engineering, 2023, 35(2): 1296-1310.
[13]
NOYN F, MCGUINNESSD L. Ontology development 101: A guide to creating your first ontology[J/OL]. [2023-11-28].
NOYN F, CRUBEZYM, FERGERSONR W, et al. Protégé-2000: An open-source ontology-development and knowledge-acquisition environment[J]. AMIA Annual Symposium Proceedings AMIA Symposium, 2003, 2003: 953.
[19]
TAOY, LIY, WUZ H. Revisiting graph neural networks for node classification in heterogeneous graphs[C]//IEEE International Conference on Multimedia and Expo (ICME). Shenzhen:IEEE, 2021: 1-6.
[20]
LYUQ S, DINGM, LIUQ, et al. Are we really making much progress?: Revisiting, benchmarking and refining heterogeneous graph neural networks[C]//Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. Virtual: ACM, 2021: 1150-1160.
[21]
WANGX, JIH Y, SHIC, et al. Heterogeneous graph attention network[C]//The World Wide Web Conference. San Francisco:ACM, 2019: 2022-2032.
FORMANG. An extensive empirical study of feature selection metrics for text classification[J]. Journal of Machine Learning Research, 2003, 3: 1289-1305.