Analysis of Fall-from-Height Accidents in Construction Based on Text Mining Technology and Improved Apriori Algorithm

Abstract

In recent years, fall-from-height accidents have frequently occurred in construction activities, posing severe risks to workers’ safety and impeding the sustainable development of construction enterprises as well as social stability. Due to the complexity and multifactorial nature of such accidents, traditional safety risk assessment methods face significant limitations in uncovering their underlying causes. To address this issue, this study develops a novel analytical framework that integrates text mining with an improved Apriori algorithm. A standardized text preprocessing pipeline is established, including data collection, construction of a domain-specific lexicon, and synonym-based term unification. Key features are extracted using the TF-IDF method, while thematic patterns are identified through LDA topic modeling. To overcome the contextual insensitivity of conventional association rule mining, the Apriori algorithm is enhanced by introducing time-based constraints, enabling the discovery of accident causation patterns that differ between daytime and nighttime. Using 1064 accident reports from 22 provinces in China, the framework extracted 40 high-frequency accident-causing features and generated a richer set of meaningful association rules compared to the standard algorithm. The results indicate that insufficient safety protection, inadequate worker training, and management deficiencies are the predominant causes of fall-from-height accidents. Building on these insights, the study proposes targeted preventive measures. The findings make significant theoretical contributions by enhancing methodological frameworks for accident analysis while also providing practical insights to improve safety management practices in the construction industry.