Large-scale and Deep Spatiotemporal Point-Process Models

• NCJ Number: 310347
• Author(s): Yuan, Baichuan
• Date Published: 2020
• Length: 136 pages
• Grant Number(s): 2018-R2-CX-0013
• Agencies: NIJ-Sponsored
• Publication Type: Thesis/Dissertation

Annotation

This paper explores the development of large-scale and deep spatiotemporal point-process models.

Abstract

In this thesis, researchers develop new multivariate models with generalization ability and scalability. The authors develop a nonparametric method for multivariate spatiotemporal Hawkes processes with applications on network reconstruction. The results demonstrate that, in comparison to using only temporal data, their approach yields improved network reconstruction, providing a basis for meaningful subsequent analysis—such as examinations of community structure and motifs—of the reconstructed networks. The authors present a fast and accurate estimation method for multivariate Hawkes processes. This method, with guaranteed consistency, combines two estimation approaches. Extensive numerical experiments, with synthetic data and real-world social network data, show that this method improves the accuracy, scalability, and computational efficiency of prevailing estimation approaches. Moreover, it greatly boosts the performance of Hawkes process-based models on social network reconstruction and helps to understand the spatiotemporal triggering dynamics over social media. The authors focus on multivariate spatial point processes, which can describe heterotopic data over space, introducing a declustering-based hidden-variable model that leads to an efficient inference via a variational autoencoder (VAE). The authors also prove that this model is a generalization of the VAE-based model for collaborative filtering. This leads to an interesting application of spatial point-process models to recommender systems. Experimental results show the method's utility on both synthetic data and real-world data. Finally, the authors show how multivariate point processes can be applied to opioid overdose events and real-time prediction of the hourly crime rate.