Abstract:Information popularity prediction is important yet challenging in various domains, including viral marketing and news recommendations. The key to accurately predicting information popularity lies in subtly modeling the underlying temporal information diffusion process behind observed events of an information cascade, such as the retweets of a tweet. To this end, most existing methods either adopt recurrent networks to capture the temporal dynamics from the first to the last observed event or develop a statistical model based on self-exciting point processes to make predictions. However, information diffusion is intrinsically a complex continuous-time process with irregularly observed discrete events, which is oversimplified using recurrent networks as they fail to capture the irregular time intervals between events, or using self-exciting point processes as they lack flexibility to capture the complex diffusion process. Against this background, we propose ConCat, modeling the Continuous-time dynamics of Cascades for information popularity prediction. On the one hand, it leverages neural Ordinary Differential Equations (ODEs) to model irregular events of a cascade in continuous time based on the cascade graph and sequential event information. On the other hand, it considers cascade events as neural temporal point processes (TPPs) parameterized by a conditional intensity function which can also benefit the popularity prediction task. We conduct extensive experiments to evaluate ConCat on three real-world datasets. Results show that ConCat achieves superior performance compared to state-of-the-art baselines, yielding a 2.3%-33.2% improvement over the best-performing baselines across the three datasets.

1 Model Overview Observed Event Sequences Feature Extraction Feature Driven Discriminator MLE-Based Predictor Interpreter

Adversarial Training Model Unifying Feature Driven and Point Process Perspectives for Event Popularity Prediction.

A Real-Time Method to Predict Social Media Popularity

Feature Driven and Point Process Approaches for Popularity Prediction

Leveraging Adaptive Peeking Window to Improve Self-Exciting Point Process Model for Popularity Prediction

PREP: Pre-training with Temporal Elapse Inference for Popularity Prediction

EPAB: Early Pattern Aware Bayesian Model for Social Content Popularity Prediction

BEEP: a Bayesian perspective Early stage Event Prediction model for online social networks

A Feature Generalization Framework for Social Media Popularity Prediction

Predicting Popularity Via a Generative Model with Adaptive Peeking Window

MM-Pred: A Deep Predictive Model for Multi-attribute Event Sequence

On Your Mark, Get Set, Predict! Modeling Continuous-Time Dynamics of Cascades for Information Popularity Prediction

Predicting Key Events in the Popularity Evolution of Online Information

MM-CPred: A Multi-task Predictive Model for Continuous-Time Event Sequences with Mixture Learning Losses

Sequential Prediction of Social Media Popularity with Deep Temporal Context Networks

Multi-Modal Video Feature Extraction for Popularity Prediction

DeepBlue: Bi-layered LSTM for Tweet Popularity Estimation

Predicting Hot Events in the Early Period Through Bayesian Model for Social Networks

INITIATOR: Noise-contrastive Estimation for Marked Temporal Point Process

Predicting future popularity trend of events in microblogging platforms

Model Can Be Subtle: Two Important Mechanisms for Social Media Popularity Prediction