Flash crowd prediction in Twitter

Computers have made their way into all realms of human actions. The emergence of high-speed networks and the popularization of mobile devices resulted in the increase of the number of accesses to online content on the Internet. Powerful servers have become inevitable as they support huge data processing. Flash crowd occurs due to popular content which attracts large crowd that demands immediate and instantaneous service which the server(s) find challenging. Research solutions till date propose dynamic substitution of resources or server(s) which is not cost effective. Our objective is to predict popular content that addresses the bottleneck's caused to server(s) by flash crowd. We develop a generic model that tracks trending topics in Twitter by extracting features through sentiment analysis, sarcasm analysis, trend analysis, emotional divergence, hash tag processing and predicts flash crowd through Naïve bayes classifier We benchmark the precision and recall of Naïve bayes classifier with KNN classifier to gauge its performance and throughput. The model depicts infectious rate which in turn helps in Flash crowd prediction in real time.