Research And Implementation Of Community Division In Social Network Based On Parallel Graph Computing

The significance of using parallel graphs in large-scale complex networks lies in the fact that the integration of parallel computing,graphs,and community partitioning algorithms makes the social network community more precise,faster,and more reliable.This study analyzes community behavior and community sentiment analysis.There are wide practical applications of recommendation systems,advertisement targeting,and social stability.By adopting improved spectral clustering algorithm and implementing parallelization,the similarity between social network nodes is calculated by using a trigonometric model in the improved algorithm,thereby changing the construction process of the spectral clustering adjacency matrix.Finally,the efficiency of the algorithm for evaluating the evaluation index of the community division index was used to measure the efficiency of the algorithm,and the algorithm was compared with the current better community partitioning algorithm.The comparison shows that the algorithm is efficient,accurate,and scalable.The research shows that the community is characterized by a high degree of closeness in the community and a low level of closeness among the communities.In the large-scale network map,there are still many overlapping areas of the community after the community is divided.Taking into account these factors may lead to accurate community division.To reduce the degree,a sub-graph partitioning-based clustering algorithm was introduced to detect overlapping communities,so as to optimize the accuracy of the improved spectral clustering algorithm in the community division and combine the parallel calculation methods.The above method was further used to further confirm the efficiency of the proposed method,accuracy,reliability characteristics.The experiment is carried out by combining the large-scale dataset Twitter dataset and the Stanford Large Network Dataset Collection test dataset.The experimental results show that the proposed method is scalable and can quickly and accurately divide the large-scale complex network into communities.