RC-Tweet: Modeling and predicting the popularity of tweets through the dynamics of a capacitor

摘要

A novel model for the popularity evolution of tweets - the RC-Tweet - is introduced. Its originality stems from the revelation of the astounding similarity between the popularity growth of tweets and the charging dynamics of a capacitor in an RC-Circuit. Fitting the model to empirical retweet patterns pertaining to cascades of all sizes, accurate goodness-of-fit statistics are obtained. These findings illustrate that the RCTweet precisely captures the dynamics of retweets, through a macroscopic mechanistic model comprising just two independent parameters. Exploring the predictive power of the model on retweet cascades of various sizes, it was found that accurate popularity forecasts, with a 6.95% average error, are produced using only the posting time of retweets, mostly occurring within an extremely short observation period (similar to 3min). The RC-Tweet model outperforms existing state-of-the-art popularity models in precision, speed and simplicity. It is suitable for real time predictions with minimal publicly available information, without training on existing data. The novelty and performance of the RC-Tweet model open up new opportunities for theoretical and applied research in the popularity evolution mechanisms in social media. Also, the RC-Tweet model is potentially applicable to the explanation and prediction of the popularity dynamics in other fields, such as commerce, financial markets, fashions and trends. (C) 2020 Elsevier Ltd. All rights reserved.