Modeling Events with Cascades of Poisson Processes

TL;DR

This paper introduces a probabilistic model for continuous-time events where each event triggers a Poisson process of subsequent events, enabling scalable inference for large datasets like Twitter and Wikipedia revisions.

Contribution

The paper proposes a new superposition-based Poisson process model with an EM algorithm for efficient, distributed inference on large-scale event data.

Findings

Effective modeling of Twitter messages and Wikipedia revisions

Scalable EM algorithm for large datasets

Superposition of Poisson processes captures event cascades

Abstract

We present a probabilistic model of events in continuous time in which each event triggers a Poisson process of successor events. The ensemble of observed events is thereby modeled as a superposition of Poisson processes. Efficient inference is feasible under this model with an EM algorithm. Moreover, the EM algorithm can be implemented as a distributed algorithm, permitting the model to be applied to very large datasets. We apply these techniques to the modeling of Twitter messages and the revision history of Wikipedia.