Investigation of Critical Factors for Future-Proofed Transportation Infrastructure Planning Using Topic Modeling and Association Rule Mining

Most existing studies on transportation infrastructure planning focus on only one or a few critical factors. In addition, the interrelationships among different planning factors were seldom investigated. Therefore, this study aims to develop a holistic understanding of various critical factors and their interrelationships toward future-proofed transportation infrastructure planning. A novel text mining-based approach was proposed in this study to identify the critical factors and their interrelationships based on selected transportation infrastructure planning publications. Two topic modeling techniques, i.e., latent Dirichlet allocation (LDA) and nonnegative matrix factorization (NMF), were used to identify the critical and emerging topics that may affect transportation infrastructures, resulting in the automatic identification of critical factors. These factors were compiled and converted to a four-level taxonomy via bottom-up grouping. Association rule mining (ARM) was then used to discover relations among the identified factors. Among these interrelationships, eight were found to be significant based on confidence and lift values as two quantitative measures of association rules. These findings could guide transportation infrastructure planners and decision makers to have a holistic approach to planning, building, and managing our transportation infrastructure in the face of future risks and opportunities. This study also demonstrates the potential of using text mining techniques to explore new knowledge in civil infrastructure planning.