Efficient Frequent Subgraph Mining for Dynamic Network Graphs Using Golden Dung Graph Hybridization

Abstract

Frequent subgraph mining (FSM) is one of the most critical procedures for mining meaningful patterns in large and dynamic graph datasets, common in several applications, such as social networks and biological data analysis. Traditional FSM methods are developed primarily with static graphs in mind and, thus, are inefficient when applied to dynamic data, especially data that updates continuously. This paper provides a novel framework of efficient FSM for dynamic network graphs with the support of a four-phase approach involving preprocessing, map, shuffle, and sort, and reduce phases. The hybrid optimization approach developed is known as Golden Dung Graph Hybridization (GDGH) and is a synchronization of Dung Beetle Optimization Algorithm and Golden Jackal Optimization Algorithm to optimize subgraph selection. For subgraph embedding and isomorphism testing, we further conduct a comparative study of several message-passing neural networks. Furthermore, this study conducts extensive experiments on several datasets that show significant superiority over the existing FSM methods in processing time, memory efficiency, and accuracy to demonstrate the efficacy of the proposed framework.