Revealing non-work travel patterns under the influence of multiple actors: An integrated geospatial agent-based approach

Abstract

Activity-based modeling is inherently complex, encompassing a variety of assumptions, actors, and emergent patterns. Uncovering underlying uncertainties requires integrating a diverse range of models and actors, and examining different assumptions implemented throughout the modeling process. Agent-Based Models (ABMs), are recognized as powerful tools for addressing the complexities inherent in activity-based modeling. Although the advantages of ABMs in activity-based models have been demonstrated, there remains a need for further exploration in two key areas: the application of integrated geospatial models to define spatial actors in ABMs, and the utilization of pattern recognition approaches, including Global Moran's I and Local Indicators of Spatial Association (LISA), to uncover the hidden mechanisms of activity-related commuting. This paper proposes an integrated Geospatial-ABM approach to address this gap. The Geospatial-ABM was validated locally using real-world data and Error percent procedure to ensure its robustness. The model validation resulted in an average percentage error of 8.8 %, showcasing its strong potential for generalization. A case study was conducted for Zanjan, Iran, using data on trip generation, built-environment characteristics, and surveyed travel behavior. The initial findings from the descriptive statistics of the simulation results reveal that shopping destinations are predominant in intra-urban commutes, with nearly half of visits to urban amenities being attributed to shopping centers. Subsequent results from pattern recognition analysis, environmental, and office-bank travels indicate, concentrating around the city-center and leading to a temporary surge in daytime traffic volume in the city center.