A power-law decay model with autocorrelation for posting data to social networking services

TL;DR

This paper introduces a power-law decay model with autocorrelation to accurately describe posting activity on social networks around major events, capturing interest fluctuations before and after the events.

Contribution

The paper presents a novel model combining power-law decay and autocorrelation for social media posting data, improving fit to real-world data around significant events.

Findings

Model fits social media posting data well

Captures pre- and post-event interest fluctuations

Incorporates autocorrelation in decay modeling

Abstract

We propose a power-law decay model with autocorrelation for posting data to social networking services concerning particular events such as national holidays or major sport events. In these kinds of events we observe people's interest both before and after the events. In our model the number of postings has a Poisson distribution whose expected value decays as a power law. Our model also incorporates autocorrelations by autoregressive specification of the expected value. We show that our proposed model well fits the data from social networking services.