Expanding Domain Knowledge Elements for Metro Construction Safety Risk Management Using a Co-Occurrence-Based Pathfinding Approach

Abstract

Knowledge is a contribution factor leading to more effective and efficient construction safety management. Metro construction practitioners always find it difficult to determine what specialized knowledge is needed in order to lead to better safety risk management. Currently, domain knowledge elements are generally determined by experts, which is coarse-grained and uncomprehensive. Therefore, this paper aims to provide a structure of domain knowledge elements, using an automatic approach to expand domain knowledge elements (DKEs) from a big dataset of unstructured text documents. First, the co-word co-occurrence network (CCN) was used to find the connected knowledge elements, and then the association rule mining (ARM) was compiled to prune the weakly related subnetworks, leaving the strong associated elements. Finally, a list of DKEs in the metro construction safety risk management was obtained. The result shows that the obtained DKEs are more comprehensive and valuable compared to previous studies. The proposed approach provides an automatic way to expand DKEs from a small amount of known knowledge, minimizing the expert bias. This study also contributes to building a fine-grained knowledge structure for metro construction safety risk management. The structure can be used to guide safety training and help knowledge-based safety risk management.