Paper Title
Hybrid Deep Neural Network (HCNN) based Community Detection Approaches for Optimization of Overlapping Dynamic Social Network

Abstract
Abstract - Popular methods for discovering and identifying user community social patterns include Social Network Analysis (SNA). The development of effective community detection systems has become a computational problem due to the exponential growth of these networks (in terms of users and interactions). User profiles on social networking sites can be as broad or as narrow as the user desires. We may forecast and recommend relationships, behaviours, and daily activities based on the data we collect on a regular basis. Massive volumes of data are generated every day by social media platforms like Facebook, Twitter, Instagram, and so on. Graphs are a common tool for community detection[1]. SNA can better investigate social networks using multimedia data thanks to Multimedia Social Networks (MSNs). These networks aid SNA in the analysis of effects, the discovery of experts, the identification of communities, and the suggesting of items. Our solution uses Convolutional Neural Networks to simultaneously use topological and contextual information for semi-supervised community discovery. Sparse matrices are used to represent network connections in order to lower the computational cost[2]. MSN user relationships are represented in a hypergraph - based model using multimedia data. Adjacency matrix was created by looking at pathways between users with significant hyperarcs in basic hypergraph[3]. A GRU, RNN HCNN was used to create this matrix for sparse data and studied separate. We examine the influence of numerous community-based social contexts by introducing a hierarchical attention mechanism that is based on the structure of the community[4]. Second, we select the neighbours who are directly connected to us, also known as the social settings that are based on our successor networks.This study discusses community detection approaches using all variation of Neural Network based learning.Our technique surpasses other algorithms in terms of running time when tested on huge data sets[5]. Keywords - CNN; GRU; Deep Learning; Network Analysis; Community Detection;Social Computing; Graph Neural Network; Multimedia Social Networks;Convolutional Networks;