Mistikoglu, GulgunGerek, Ibrahim HalilErdis, ErcanUsmen, P. E. MumtazCakan, HulyaKazan, Emrah Esref2024-09-182024-09-1820150957-41741873-6793https://doi.org/10.1016/j.eswa.2014.10.009https://hdl.handle.net/20.500.12483/9995Data mining (DM) techniques have not been adopted on a wide scale for construction accident data analysis. The decision tree (DT) technique is a supervised data mining method that shows good promise for this purpose. The C5.0 and CHAID algorithms were employed in this study to construct decision trees and to extract rules that show the associations between the input and output variables (attributes) for roofer fall accidents. Data obtained from the US Occupational Safety and Health Administration (OSHA) was incorporated in this research. Degree of injury (fatality vs. nonfatal injury) was selected as the output attribute, and a multitude of input attributes were included in the study. Two models based on the algorithms were developed and validated. The results showed that decision trees provided specific and detailed depictions of the associations between the attributes. It was found that fatality chances increased with increasing fall distance and decreased when safety training was provided. The most important input attributes in the models were identified as the fall distance, fatality/injury cause, safety training, and construction operation prompting fall, meaning that these factors had the best predictive power related to whether a roofer fall accident would result in a fatality or nonfatal injury. (C) 2014 Elsevier Ltd. All rights reserved.eninfo:eu-repo/semantics/closedAccessFall accidentsData miningDegree of injuryDecision treePredictive powerArticle4242256226310.1016/j.eswa.2014.10.0092-s2.0-84910644907Q1WOS:000347579500043Q1