Transport simulation models play a crucial role in transportation planning, design, and operations, allowing for the replication of various scenarios through the incorporation of real-world data and parameters. Recently, agent-based transport models have gained prominence for their ability to simulate intricate metropolitan transport systems. These models take into account the distinct characteristics, decision-making processes, and interactions of individual agents. Among the array of agent-based transport models, MATSim stands out as a potent and adaptable tool for modeling transportation systems. A critical aspect of MATSim’s input preparation involves assigning activity location points using land use raster data. However, the characteristics of land use raster data present limitations in certain urban case studies such as Stockholm. In response, some researchers have turned their attention to buildings shapefile data, a commonly used geospatial data format. This study aims to improve the activity location assignment model by developing an evaluation workflow of model uncertainty for different geospatial input data in MATSim and empirically analyzing their impacts on simulation outcomes. Despite acknowledging data availability and activity representation limitations, the study’s results demonstrate that utilizingbuildings shapefiles as input data yields more consistent outcomes with reduced uncertainty. This suggests the promising potential of buildings shapefiles as a favorable data source for transportation modeling and planning within the studied scenarios.