MultiRepast4py: A Framework for Agent-Based Simulations on Multilayer Networks

Keng-Lien Lin, Parinaz Naghizadeh

Keywords: Agent Based Simulations, Multilayer Networks, Repast

TL;DR: We develop a framework that enables agent-based simulations on multilayer networks (i.e., networks of networks), and showcase it using simulation examples on rumor spreading.

Abstract:

Agent-Based Simulations (ABS) offer a powerful approach for analyzing how individual agents' decisions and interactions within networked systems lead to system outcomes. ABS have been widely used across various fields, including in the study of the spread of diseases and (mis)information. However, traditional ABS platforms, such as NetLogo and Repast, simplify models by assuming a single network of interactions between agents. In reality, agents' interactions are typically multi-layered (i.e., involve multiple interconnected networks that influence agents' decisions). To address this limitation, we developed $\texttt{MultiRepast4py}$, a multilayer simulation tool extending the simulation capabilities of Repast4py.

Our framework enables simulations on multilayer networked systems by efficiently reconstructing network data and utilizing agent attributes, allowing agents to dynamically access multilayer connections during simulation. By maintaining Repast4py’s scalability and minimizing memory overhead, $\texttt{MultiRepast4py}$ ensures high performance for large-scale simulations. Through simulation examples on the spread of information in social networks, we showcase how $\texttt{MultiRepast4py}$ can enable more comprehensive agent-based simulations, guiding improved predictions and interventions.

Submission Number: 17