Modeling and predicting popularity dynamics via reinforced poisson processes

H. Shen, D. Wang, C. Song, A.-L. Barabási
Proceedings of the 28th AAAI Conference on Artificial Intelligence
291-297 (2014)
January 4, 2014


An ability to  predict the popularity dynamics of individual items within a complex evolving  system has important implications in an array of areas. Here we propose a  generative probabilistic framework using a reinforced Poisson process to  explicitly model the process through which individual items gain their  popularity. This model distinguishes itself from existing models via its  capability of modeling the arrival process of popularity and its remarkable  power at predicting the popularity of individual items. It possesses the  flexibility of applying Bayesian treatment to further improve the predictive  power using a conjugate prior. Extensive experiments on a longitudinal  citation dataset demonstrate that this model consistently outperforms  existing popularity prediction methods.

Related publications