BIG DATA OVER A DETECTED COMMUNITY
International Research Journal of Modernization in Engineering Technology and Science · 2023
Network analysis relies largely on discovering similarity across communities, whilst a community can be strengthened with the help of content information.However, it is largely inhibited by noise that is present in most networks, especially in the link structure.This paper presents a basic approach to combine content with link information in graph-based structures to assist community discoveries.It also tries to reduce the impact of noises commonly found in social networking sites as well as Web-based information networks.We propose to calculate strength of a signal between nodes across the network by combining the link strength, which denotes the probability of that link lying inside a community, with similar content that can be estimated using cosine similarity, or the Jaccard coefficient.Furthermore, we discuss an edge-sampling process to retain locally-relevant edges for every node of the graph.The graph that results could be then clustered by using standard algorithms for community discoveries, such as Markov-clustering and METIS.We experimented on real-world datasets (Wikipedia, CiteSeer and Flickr) by changing sizes and parameters in order to understand the efficacy of our approach versus existing ones.We tried to find a beneficial method to combine approaches for a content and link analysis and a faster biased, graph-sampling approach.