Predicting the Popularity of Online Serials with Autoregressive Models

Chang, Biao; Zhu, Hengshu; Ge, Yong; Chen, Enhong; Xiong, Hui; Tan, Chang Wei

doi:10.1145/2661829.2662055

Cited by 27 publications

(17 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Prediction performances are further improved when adding features which account for specific contexts of the information diffusion. For example, Li et al [22] add propagation network measurements, Chang et al [6] add sophisticated features relating to audience behavior when predicting the popularity of serials and Vallet et al [38] construct features relating to two social networks: Youtube and Twitter. Many proposed features cannot be used outside of the proposed context, being either data-source specific or relying on information that is not publicly accessible.…”

Section: Related Workmentioning

confidence: 99%

Feature Driven and Point Process Approaches for Popularity Prediction

Mishra

Rizoiu

Xie

2016

Proceedings of the 25th ACM International on Conference on Information and Knowledge Management

159

168

View full text Add to dashboard Cite

Predicting popularity, or the total volume of information outbreaks, is an important subproblem for understanding collective behavior in networks. Each of the two main types of recent approaches to the problem, feature-driven and generative models, have desired qualities and clear limitations. This paper bridges the gap between these solutions with a new hybrid approach and a new performance benchmark. We model each social cascade with a marked Hawkes self-exciting point process, and estimate the content virality, memory decay, and user influence. We then learn a predictive layer for popularity prediction using a collection of cascade history. To our surprise, Hawkes process with a predictive overlay outperform recent feature-driven and generative approaches on existing tweet data [43] and a new public benchmark on news tweets. We also found that a basic set of user features and event time summary statistics performs competitively in both classification and regression tasks, and that adding point process information to the feature set further improves predictions. From these observations, we argue that future work on popularity prediction should compare across feature-driven and generative modeling approaches in both classification and regression tasks

show abstract

Section: Related Workmentioning

confidence: 99%

Feature Driven and Point Process Approaches for Popularity Prediction

Mishra

Rizoiu

Xie

2016

Proceedings of the 25th ACM International on Conference on Information and Knowledge Management

159

168

View full text Add to dashboard Cite

show abstract

“…Computing the final size in the deterministic model is straightforward, as it results directly from the differential equations in Eq. (3)- (5). Allen [1] shows that dividing Eq.…”

Section: H4 Estimating N In the Deterministic Sirmentioning

confidence: 99%

SIR-Hawkes

Rizoiu

Mishra

Kong

et al. 2018

Proceedings of the 2018 World Wide Web Conference on World Wide Web - WWW '18

View full text Add to dashboard Cite

Among the statistical tools for online information diffusion modeling, both epidemic models and Hawkes point processes are popular choices. The former originate from epidemiology, and consider information as a viral contagion which spreads into a population of online users. The latter have roots in geophysics and finance, view individual actions as discrete events in continuous time, and modulate the rate of events according to the self-exciting nature of event sequences. Here, we establish a novel connection between these two frameworks. Namely, the rate of events in an extended Hawkes model is identical to the rate of new infections in the Susceptible-Infected-Recovered (SIR) model after marginalizing out recovery events -which are unobserved in a Hawkes process. This result paves the way to apply tools developed for SIR to Hawkes, and vice versa. It also leads to HawkesN, a generalization of the Hawkes model which accounts for a finite population size. Finally, we derive the distribution of cascade sizes for HawkesN, inspired by methods in stochastic SIR. Such distributions provide nuanced explanations to the general unpredictability of popularity: the distribution for diffusion cascade sizes tends to have two modes, one corresponding to large cascade sizes and another one around zero.We fit the parameters of the HawkesN model to observed data by maximizing the log-likelihood function Eq. (28). We use AMPL, an industry standard for modeling optimization problems and with a transparent interfaces to powerful solvers. We start with an introduction of AMPL (Sec. B.1), we describe our optimization setup and the employed solvers (Sec. B.2) and we finish with the R interface that we constructed for AMPL (Sec. B.3).

show abstract

“…Researchers have tried to model and predict how the amount of attention will be devoted over time to a given piece of online information (that is, popularity evolution). Our work relates to the two directions of popularity evolution: popularity evolution patterns and popularity evolution prediction [23, 24]. …”

Section: Background and Related Workmentioning

confidence: 99%

Predicting Key Events in the Popularity Evolution of Online Information

Hu¹,

Hu²,

Fu³

et al. 2017

PLoS ONE

View full text Add to dashboard Cite

The popularity of online information generally experiences a rising and falling evolution. This paper considers the “burst”, “peak”, and “fade” key events together as a representative summary of popularity evolution. We propose a novel prediction task—predicting when popularity undergoes these key events. It is of great importance to know when these three key events occur, because doing so helps recommendation systems, online marketing, and containment of rumors. However, it is very challenging to solve this new prediction task due to two issues. First, popularity evolution has high variation and can follow various patterns, so how can we identify “burst”, “peak”, and “fade” in different patterns of popularity evolution? Second, these events usually occur in a very short time, so how can we accurately yet promptly predict them? In this paper we address these two issues. To handle the first one, we use a simple moving average to smooth variation, and then a universal method is presented for different patterns to identify the key events in popularity evolution. To deal with the second one, we extract different types of features that may have an impact on the key events, and then a correlation analysis is conducted in the feature selection step to remove irrelevant and redundant features. The remaining features are used to train a machine learning model. The feature selection step improves prediction accuracy, and in order to emphasize prediction promptness, we design a new evaluation metric which considers both accuracy and promptness to evaluate our prediction task. Experimental and comparative results show the superiority of our prediction solution.

show abstract

Predicting the Popularity of Online Serials with Autoregressive Models

Cited by 27 publications

References 31 publications

Feature Driven and Point Process Approaches for Popularity Prediction

Feature Driven and Point Process Approaches for Popularity Prediction

SIR-Hawkes

Predicting Key Events in the Popularity Evolution of Online Information

Contact Info

Product

Resources

About