Dependency Tracking for Relation-Based Modeling. Using Attribute-Access Tracking to Enable Efficient Execution and Support Analysis of Simulation Traces in Relation-Based Modeling.

Jankowski, Miłosz (2025) Dependency Tracking for Relation-Based Modeling. Using Attribute-Access Tracking to Enable Efficient Execution and Support Analysis of Simulation Traces in Relation-Based Modeling. Bachelor thesis, Institute for Visual and Analytic Computing, University of Rostock.

Text Bachelorarbeit_MiloszJankowski.pdf - Published Version Restricted to Registered users only Download (3MB)

Abstract

Relation-based modeling (RBM) represents a paradigm shift from traditional agent-based approaches, emphasizing dynamic relations and assemblages over individual entities, and providing unique insights into a modeled social-ecological system. However, RBM faces two challenges: 1) the existing DSL (RBML) only implements an inefficient simulation algorithm, and 2) causal analysis of RBMs is labor-intensive, requiring manual tracing of assemblage compositions over time. To address these challenges, we devise and implement a dependency tracking mechanism within RBML and use it to implement the more efficient Next Reaction Method, achieving a factor-two speedup. Furthermore, we attempt to use the tracking mechanism and the NRM implementation to support analysis of RBMs, but discover that this method is not suitable for analysis of this class of models. We instead propose a prototypical analysis method, which automates parts of the analysis process, allowing zooming in to particularly interesting parts of an event log. We showcase the practicality of the proposed analysis method on a model of a small-scale Mexican fishery.