Initializing Agent-Based Models With Clustering Archetypes

Keywords

Agent-based models; GitHub archetypes; Stable clustering; Unsupervised learning

Abstract

Agent-based models are a powerful tool for predicting population level behaviors; however their performance can be sensitive to the initial simulation conditions. This paper introduces a procedure for leveraging large datasets to initialize agent-based simulations in which the population is abstracted into a set of archetypes. We show that these archetypes can be discovered using clustering and evaluate the benefits of selecting clusters based on their stability over time. Our experiments on the GitHub dataset demonstrate that simulation runs performed with the clustering archetypes are more successful at predicting large-scale activity patterns.

Publication Date

1-1-2018

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Volume

10899 LNCS

Number of Pages

233-239

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

DOI Link

https://doi.org/10.1007/978-3-319-93372-6_27

Socpus ID

85049777151 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85049777151

This document is currently not available here.

Share

COinS