Modeling Morphology of Cascades in Online Social Networks using Multi-Order Markov Chains

Cascades represent an important phenomenon across various disciplines such as sociology, economy, marketing, and epidemiology. An important property of cascades is their morphology, which encompasses their structure, shape, and size. However, we believe that cascade morphology has not been thoroughly studied in prior literature. The goal of this paper is to develop a model that allows us to quantitatively and rigorously analyze, characterize, and classify cascade morphology. Towards this end, we propose M4, a Multi-order Markov Model for the Morphology of cascades in online social networks. M4 can represent cascades with arbitrary structures, shapes, and sizes, and also allows us to classify cascades based on their underlying attributes. M4 essentially provides a quantitative tool for analyzing and classifying cascades solely based on their morphology. For validation, we apply M4 model to solve the following cascade size prediction problem: given the first τ1 edges in a cascade, can we predict if the cascade will have a total of more than τ2 edges over its lifetime, where τ2 > τ1? The results of our experiments, conducted on a Twitter dataset, show that M4 achieves classification accuracy of up to 91.2% for this prediction problem.