Comprehensive causal analysis of occupational accidents’ severity in the chemical industries; A field study based on feature selection and multiple linear regression techniques

• Published in: Bihdāsht va īmanī-i kār Vol. 9; no. 4; pp. 298 - 310
• Main Authors: Ahmad Soltanzadeh, Hamidreza Heidari, Heidar Mohammad, Abolfazl Mohammadbeigi, Vali Sarsangi, Milad Darakhshan Jazari
• Format: Journal Article
• Language: Persian
• Published: Tehran University of Medical Sciences 01.12.2019
• Subjects: accident severity rate (asr); chemical industry; features selection; multiple linear regression; occupational accidents
• ISSN: 2251-807X; 2383-2088

Introduction: The causal analysis of occupational accidents’ severity in the chemical industries may improve safety design programs in these industries. This comprehensive study was implemented to analyze the factors affecting occupational accidents’ severity in the chemical industries. Methods and Materials: An analytical study was conducted in 22 chemical industries during 2016-2017. The study data included 41 independent factors and 872 accidents in a ten-year period (2006-2015) as a dependent variable. Feature selection algorithm and multiplied linear regression techniques were used to analyze this study. Results: Accident severity rate mean was calculated 214.63 ± 145.12. The results of feature selection showed that 30 factors had high impacts on the severity of accidents. In addition, based on regression analysis, the severity of accidents in the chemical industries was affected by 22 individuals, organizational, HSE training, risk management, unsafe conditions and unsafe acts, as well as accident types (p<0.05). Conclusion: The findings of this study confirmed that accidents’ severity in the chemical industry followed the multi-factorial theory. In addition, the main finding of this study indicated that the combination of features selection algorithm and multiple linear regression methods can be useful and applicable for comprehensive analysis of accidents and other HSE data.