A New Popularity Prediction Model for Detection of Trending Topics based on Support Vector Machine over Online Social Networking

Abstract

Social network is the most reliable platform to share data. It produces a huge number of subjects each day. Online media is critical for business associations to learn the popularity of these themes as fast as could be expected. This paper is centered around the investigation of five distinctive existing trending detection methods, such as The Prediction Model based on SVM, Multi Linear Model, Dual Attention Model, Self Attention based Model, Trust Prediction Model, etc. But a few upsides and downsides are seen in these methods. These methodologies have been analyzed to address the limits of the trending topic popularity prediction such as video lifetime, burst and evolution pattern, execution, error rate, time, overhead, throughput, delay, limit, and so on. In this paper, “Rapid Miner” tool is used to implement the proposed method. The correlation matrix is used for pattern detection, then clustering is performed over the output of the correlation matrix and obtains the data in cluster format i.e. in similar sequence, and again random clustering technique is performed over previous K-means clustering result, in which, it simply divides topic information into time slices which are utilized as a unit, and such units are provides to the two unique detectors, for example, lifetime and bursty detector. At last, SVM classification predicts the outcomes in the type of a "Trending Topic" and a "Non-Trending Topic". The proposed strategy is “Detection of the Trending Topic” which dependent on SVM. The proposed method recognizes the lifetime of recordings, bustiness of recordings, and evolution pattern. It decreases the mistake rate, time, overhead, delay, and enhances the precision, execution, use of mistake rate as a measurement, throughput, ability to improve the trending subject identification rate, etc. This proposed strategy concentrates on a trending subject and fake data over OSN.