A Machine Learning Study on the Role of Behavioral and Demographic Factors in Mining Injuries

Abstract

Predictive models for work-related injuries in mines are complicated, with personal, technical, and societal aspects all playing a role. Numerous studies have reported that human behavior is a significant factor in mine injuries. Many researchers have opined that maintaining a high level of safe human behavior is a challenging task. In Indian mines, the risk management team uses hierarchy control charts to avoid incidents, but managing human behavioral factors which lead to injury is more challenging. Owing to these constraints, we attempted to find the significant human behavioral factors that lead to injury in mines. In this work, we try to develop predictive models using behavioral and demographic factors to predict mine workers’ injury status and evaluate the most significant factor. To investigate the relevance of behavioral characteristics in work-related injuries, we employed three machine learning models (neural networks, random forest, and K-NN) as well as two classic statistical approaches (linear and logistic regression) to compare their results and see which method performs better. We have collected data of six factors and injury status of 186 workers through a questionnaire survey. After data analysis, a detailed comparison between machine learning models and traditional methods shows that random forest performs better than other models. This study also found that demographic features and job dissatisfaction attitude are the leading influencing factors to work-related injuries. This work is beneficial to mine management in the present Indian mining scenario to devise appropriate countermeasures.