On Quantifying Predictability in Online Social Media Cascades Using Entropy

Abstract

Predicting cascade volumes in social media communication is an important topic in furthering the use of social media for viral marketing, impact of political campaigns and in home-land security. Several techniques have been reported in the literature to estimate the cascade volumes. These algorithms use a variety of information such as Content, Structural and Temporal features, depending on their availability. Due to the spread of information infused into the algorithms the prediction accuracy has been shown in the literature to be different for different algorithms. Entropy based measures that are tailored for the differing situations of information availability have been successfully applied in the prediction scenarios in many fields including network traffic, human mobility and radio spectrum state dynamics as well as in atmospheric science. In this paper we adopt a multitude of entropy based measures for quantifying the predictability of cascade volumes in online social media communications. The limit derived from the entropy measures discussed in this paper has also been used to explain the difference in accuracies of some of the algorithms for cascade volume predictions reported in the literature. For the purpose of illustration and to demonstrate the utility of the entropy based predictability limits we have used two data sets, the MemeTracker dataset and Twitter Hashtags dataset. The results obtained in this paper demonstrate clearly the utility of entropy based measures for quantifying the predictability in online social media cascades. We have also shown that temporal relevancy is a dominant contributing factor in cascade predictability and how additional features such as the knowledge of a small number of large media sites and blogs can have significant influence on the prediction performance.