Header menu link for other important links
X
Ternary-based feature level extraction for anomaly detection in semantic graphs: an optimal feature selection basis
Published in Springer
2021
Volume: 46
   
Issue: 1
Abstract
Nowadays, the homeland security field faces more difficulties in identifying suspicious or abnormal entities in huge datasets. Even though there are numerous technologies available, the objective of finding out the anomalous instances in huge semantic graphs is still a challenging point. This is because the nodes are strongly linked with innumerable links. When a node carries unique or abnormal semantics in the network, it is considered as an abnormal node. In order to understand this idea, the semantic profile of each node is generated by modeling the graph structure using various kinds of nodes and links linked to the node at a particular distance through edges. Here, the relation between the nodes is represented by a certain weight. After framing the graph structure, ternary-based feature level extraction based on assigned weight takes place. Further, the optimal feature selection from extracted rules is deployed where second-order mutual information is assigned as the fitness or the objective function. Here, the abnormal paired nodes are optimally selected using the improved Dragonfly Algorithm (DA) based on the maximum mutual information. Since the weighting factors utilized in DA are based on fitness function, the proposed algorithm is termed as Fitness-Weighed Dragonfly Algorithm (FW-DA). The effectiveness of the proposed algorithm is substantiated by comparing it with the conventional models through various performance analyses. © 2021, Indian Academy of Sciences.
About the journal
JournalData powered by TypesetSadhana - Academy Proceedings in Engineering Sciences
PublisherData powered by TypesetSpringer
ISSN02562499