Transport infrastructure is a fundamental component of the whole infrastructure system and socio-economic development. However, it is still a challenge to identify factors affecting large-scale infrastructure due to the lack of high-quality data and inconsistent methods. To address this issue, this study developed a comparison of data- and knowledge-driven approaches in exploring factors affecting road infrastructure performance in Western Australia using network-level high-resolution road defects data. In data-driven analysis, an optimal parameters-based geographical detectors (OPGD) model, developed based on spatial heterogeneity, was developed to investigate the contributions of explanatory factors from a spatial data perspective. In knowledge-driven analysis, a questionnaire survey was performed through group interviews with regional road management teams to analyze potential explanatory factors. A spatial analytic hierarchy process (S-AHP) approach was implemented to quantify the contributions of factors based on the survey. Finally, the consistency and difference between data- and knowledge-driven approaches are evaluated based on contributions of factors and the predictions of defect risks across the road network. The results indicate that the contributions of factors tend to be similar in both approaches, and the spatial distributions of defect risks predicted by both approaches are highly correlated. The factors and risks analyzed using both methods in rural areas are more consistent than those in urban areas due to the complexity and uncertainty of road defects. Findings from this study are critical to understanding data- and knowledge-driven approaches in transport infrastructure management and determining reasonable approaches for decision-making.