Deep learning algorithms for analyzing social network influencers

Abstract

Recognizing influential users is critical in large networks because of its diverse fields. Conventional centrality techniques are often based on topographical network architectures, with different centrality methods taking into account various structural features associated to functional relevance. In several contexts, however, there is always a complicated and nonlinear relationship between a node's functional importance and its many properties, such as local location, worldwide location, and so on, which is difficult to characterize by a single centrality. This study offers a method that is based on machine learning to quantify the relevance of vertices in the scenarios of propagation in order to address this problem. When it comes to supervised learning models, such as KNN, DT, RF, SVM, and CNN ELM, distinct supervised predictors are constructed in order to identify the nodes in complex social networks that have the greatest influence on the network as a whole. In terms of accuracy, sensitivity, and specificity, as well as the F1-Score measure, the CNN ELM model that was developed performed better than alternative learning algorithms. The average F1-score that CNN ELM managed to attain was 94.2%, and its detection exactness was roughly 98.5%.