Analyzing and predicting Lifetime of trends using social networks

Abstract

Trending topics in any social networking site are those specific trends that are most popular and talked about at that point in time. The emergence, survival and eventual fading away of a trend are a function of various factors. In this paper, an exhaustive list of determinants is pioneered with respect to the survival of a trend. Using the first determinant, Interest Over Time, and utilizing sentiment analysis over the textual data associated with the trend, the longevity of a trend is predicted. Additionally, a modified and more accurate mathematical equation for Trend Survival Score is formulated where Predictive Analysis and Modeling with First Order Differential Equations is done with the other two determinants. In order to map the Lifetime Score to a timeframe relevant to real-world scenario, K-nearest neighbor approach is used (using k=3) considering how lifetime score of a trend converts to actual lifetime in the training data.