Minimization of Dynamic Rumor Influence and Propagation with Deep Learning Based Classification in Social Networks

Abstract

The practice of circulating rumor across individuals or people can be described as dissemination of information in social network. The size of the dissemination of rumors extensively differs with regard to launch nodes. The nodes are anticipated to be significant for viral marketing, if we could pick nodes that furnish to large scale diffusion. In this article, we prefer to emphasize on the negative data concerns, such as online rumors. Rumor block can well be a major disadvantage in social networks of large scale. Spiteful rumor may lead to disruption in social structure, and so should be thwarted once it is identified. In this research, the classifiers like random forest,  naïve Bayes, KNN and Deep Learning Based Boosted Ensemble are being used in order to scale down the impact of the rumor from the dataset. But several datasets are inconsistent, which means that certain data from the same section are huge in numbers, and some are infrequent. This lopsided feature of the datasets immensely impacts the efficiency of the classifier as well; hence, the SMOTE methodology is brought in to handle this concern. Additionally, Semi-Supervised Clustering Algorithm (SSCA) is employed with the details gathered through social network assessment, in order to overcome the issue of rumor circulation. Tests and trials are obtained from the endorsed large-scale global networks and authenticate the capability of the four classifiers.