Road traffic accidents pose a significant global health concern, with an alarming 1.19 million fatalities reported in 2021. Traditionally, strategies to address this challenge have relied on expert input and subjective evaluations. This study introduces an adaptive and novel two-stage model that minimizes expert interference by integrating multi-objective optimisation with association rule mining. This innovative approach provides a systematic framework to enhance the efficiency of decision-making and optimise resource allocation outcomes in road infrastructure management, facilitating adaptability to diverse objectives. A case study in Utrecht from iRAP validates the efficiency of the approach, demonstrating significant improvements across various objectives by using the non-dominant sorting genetic algorithm and enabling road local authorities to tailor investment plans to their specific road network characteristics by crash database mining. However, the methodology requires refinement, particularly in identifying risk levels considering the interactive effects of multiple road attributes. In conclusion, while representing a substantial advancement, further refinement is necessary to fully realize its potential in enhancing road safety.