Modeling the Intensity Function of Point Process Via Recurrent Neural Networks

• DOI: 10.1609/aaai.v31i1.10724
• Publication Date: 2022-06-24T18:38:34Z
• Authors: Shuai Xiao; Junchi Yan; Xiaokang Yang; Hongyuan Zha; Stephen Chu
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Statistics - Machine Learning; 0202 electrical engineering, electronic engineering, information engineering; Machine Learning (stat.ML); 02 engineering and technology; Machine Learning (cs.LG)

Event sequence, asynchronously generated with random timestamp, is ubiquitous among applications. The precise and arbitrary timestamp can carry important clues about the underlying dynamics, has lent event data fundamentally different from time-series whereby series indexed fixed equal time interval. One expressive mathematical tool for modeling point process. intensity functions of many processes involve two components: background effect by history. Due to its inherent spontaneousness, be treated as a while other need handle history events. In this paper, we model Recurrent Neural Network (RNN) units aligned indexes modeled another RNN whose are asynchronous events capture long-range dynamics. whole type prediction output layers trained end-to-end. Our approach takes an perspective process, models effect. For utility, our method allows black-box treatment which often pre-defined parametric form in processes. Meanwhile end-to-end training opens venue reusing existing rich techniques deep network process modeling. We apply predictive maintenance problem using log dataset more than 1000 ATMs global bank headquartered North America.