Tweet's popularity dynamics

Abstract

This article charts the work of a 4 month project aimed at automatically identifying patterns of tweets popularity evolution using Machine Learning and Deep Learning techniques. To apprehend both the data and the extent of the problem, a straightforward clustering algorithm based on a point to point distance is used. Then, in an attempt to refine the algorithm, various analyses especially using feature extraction techniques are conducted. Although the algorithm eventually fails to automate such a task, this exercise raises a complex but necessary issue touching on the impact of virality on social networks.