Abstract:

This study developed a method for identifying intrinsic safety evaluation indicators and the associated assessment methods based on the LDA topic model. Firstly, the definition of intrinsic safety for smart highway was proposed. Guidelines for smart highway construction from 15 provinces and municipalities were collected. The Term Frequency-Inverse Document Frequency (TF-IDF) algorithm and the LDA topic model were applied to extract topics and keywords from the corpus. By integrating relevant standards and expert evaluation, an intrinsic safety performance evaluation index system for smart highway was established. This index system included four first-level indicators: safety perception and monitoring capability, traffic operation safety assurance capability, safety decision support capability, and system and equipment reliability, which are further divided into 15 second-level indicators. The connotation of each second-level indicator was described. Subsequently, the interactive relationships among the evaluation indicators of the intrinsic safety performance of smart highway were calculated using the ANP, and the weights of each evaluation indicator were determined using the Super Decision software. The results show that data transmission reliability, electromechanical equipment operation reliability, safety risk prediction and assessment capability are key indicators affecting the intrinsic safety level of smart highways. The developed evaluation methodology enables efficient extraction of valid data from textual sources, thereby enhancing the objectivity and accuracy of intrinsic safety assessment. This approach provides actionable guidance for advancing intrinsic safety development in smart highway.

Key words: smart highway, intrinsic safety, evaluation index system, latent Dirichlet allocation (LDA) topic model, analytic network process (ANP)