BEAUT: An ExplainaBle Deep LEarning Model for Agent-Based PopUlations With Poor DaTa

作者：

Highlights：

• A deep learning simulation approach for data-poor domains is presented called BEAUT.

• The model can extract agent-based time-series patterns from a small amount of information.

• To understand the decisions made when simulating, weekly transition probability matrices are output during training to make it “explainable”.

• When compared with traditional “black box” models, BEAUT is able to outperform or perform on-par in certain evaluations.

摘要

•A deep learning simulation approach for data-poor domains is presented called BEAUT.•The model can extract agent-based time-series patterns from a small amount of information.•To understand the decisions made when simulating, weekly transition probability matrices are output during training to make it “explainable”.•When compared with traditional “black box” models, BEAUT is able to outperform or perform on-par in certain evaluations.

论文关键词：Deep q-learning,Neural fitted q-iteration,Simulation,Homelessness,Agent-based

论文评审过程：Received 26 August 2021, Revised 9 March 2022, Accepted 14 April 2022, Available online 30 April 2022, Version of Record 20 May 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108836