Towards safer mining environments: an in-depth review of predictive models for accidents

Abstract

The mining industry is of great economic significance in many nations, but it is also considered one of the most dangerous sectors due to its intrinsic characteristics. Mining accidents are a major cause of injuries and fatalities on a global scale. Therefore, this matter receives significant focus within the field of research, prompting the investigation of sophisticated algorithms and models for the analysis and prediction of mining accidents. The primary aim of these endeavors is to ascertain the key components contributing to such mishaps. The study of mining accident forecasting aims to develop technologies that provide a safer working environment and eventually contribute to preserving human lives. The primary aim of this study is to provide an in-depth overview of the latest developments in the field of mining accident prediction. This comprehensive overview spans various methodologies, encompassing time series analysis methods, statistical approaches, data science techniques, machine learning, and deep learning algorithms. Additionally, this article presents a comprehensive analysis and examination of the primary data sources commonly used to predict mining accidents. In order to analyze the material thoroughly, this paper outlines and compares the many algorithms employed to predict mining accidents. The analysis comprises an exhaustive compilation of various algorithms and a comparative evaluation. Moreover, the appropriateness of their suitability is assessed based on the characteristics of the data under analysis. The acquired outcomes and the simplicity of their interpretation and analysis are likewise subject to scrutiny. The authors have stated that the most favorable outcomes are achieved by combining two or more analytic procedures, resulting in an enhanced examination of the given results. Among the upcoming problems in mining, forecasting is expanding the scope of the proposed models and forecasts by incorporating heterogeneous data sources such as geographical data, videos, audio recordings, textual content, sentiment, and emotional intelligence.