LPCD: Incremental Approach for Dynamic Networks

Abstract

Graph mining and analytics refer to analyzing and extracting meaningful information from graph data structures. Graph mining includes techniques such as community detection, centrality analysis, and graph pattern matching to identify significant subgraphs or clusters of nodes. Graph analytics includes node classification, link prediction, and graph clustering. Graph mining and Graph analytics together are used in applications such as recommender systems, social network analysis, and fraud detection. Link Prediction and Community Detection are essential techniques in Graph mining and analytics. Link Prediction is used in real applications to recommend new friends in the social network and to predict the traffic flow in the transportation network. On the other hand, Community Detection is essential as it shares common interests or characteristics, which can help better understand social phenomena like how the community structure changes over time. One of the applications of Community Detection is transportation which uses to identify the clusters of sources and destinations that travelers frequently visit. There are issues with the traditional approaches on predicting the links and detecting the communities, and there is a demand to detect the communities incrementally. Unlike existing methods, which use the entire dataset for detecting the communities, our proposed methodology Link Prediction and Community Detection (LPCD) for dynamic networks form the links within or between the communities and can understand how the network is evolving in an incremental fashion. We experimented on benchmark datasets, and our experimental results were compared with the conventional approaches and showed better performance.