DAISY LabSharif University of Technology
Publications

Detecting Popular Social Events through Limited Observation with Deep Survival Analysis

Maryam Ramezani, Hossein Goli, AmirMohammad Izadi, Hamid R. Rabiee

arXiv preprint, 2 October 2024 Preprint

arXiv

Abstract

Identifying and analysing popular trends in social networks gives valuable insight into the dynamics of information dissemination. More importantly, by observing the dissemination pattern of a piece of information in the early stages of its expansion, it becomes possible to determine whether a cascade will become highly popular in the future. This research predicts and detects popular trends in social networks by observing limited early-stage data with a deep survival analysis based method. The proposed method is evaluated on real-world anonymised datasets from Twitter, Weibo and Digg, and is applicable to recommendation systems, reach prediction for digital content, and decision-making in digital marketing.

In short

Predicting which cascade becomes popular is usually framed as classification over a fixed observation window. Survival analysis is a better fit: the quantity of interest is when — and whether — a cascade crosses a popularity threshold, and most observations are censored because the cascade is still running.

The model is trained on early-stage observations only, and evaluated on Twitter, Weibo and Digg.

Cite this paper

@misc{ramezani2024detecting,
  title = {{Detecting Popular Social Events through Limited Observation with Deep Survival Analysis}},
  author = {Maryam Ramezani and Hossein Goli and AmirMohammad Izadi and Hamid R. Rabiee},
  year = {2024},
  eprint = {2410.01320},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2410.01320}
}

Related work