Investigation of the usage of machine learning to explore the impacts of climate change on occupational health: a systematic review and research agenda

TL;DR

This paper reviews how machine learning is used to study the effects of climate change on occupational health, identifying trends and suggesting future research directions.

Contribution

The paper provides a systematic review and research agenda for applying machine learning to understand climate change impacts on occupational health.

Findings

75% of reviewed papers used wearable sensor data with supervised machine learning techniques like SVM, RF, and KNN.

The remaining 25% used national databases for risk analysis, focusing on feature selection and classification.

The study highlights the novelty of the topic and proposes a research agenda for future exploration.

Abstract

Occupational accidents can be potentialized by factors related to the workplace or the environment, such as climatic conditions. Air temperature, wind speed, and humidity can be used to monitor occupational heat stress, leading to cramps, exhaustion, stroke, and even death. Under the climate change scenario, measuring these variables is fundamental to developing adaptation strategies for maintaining the workers’ well-being. However, when dealing with this high data volume from distinctive factors, traditional techniques are insufficient to extract all information effectively. Therefore, computational intelligence and data analytics tools can enhance data processing and analysis. Machine learning techniques have been successfully applied to occupational health and climate contexts. This paper explores the literature regarding applying these techniques to investigate the effects of climate change on occupational health. We conducted a systematic review through five scientific databases guided by three research questions, resulting in 24 selected papers. 75% of the papers screened used primary data collected from wearable sensors to monitor the well-being of workers, where we identified a trend of using supervised machine learning techniques, especially classification and regression algorithms, such as SVM, RF, and KNN. The remaining focus is on using secondary data from national databases to investigate the risk, with a trend of using feature selection techniques and classification tasks. Considering this topic is relatively new, we developed an agenda to guide future research, with suggestions to follow the trends found in this review and highlight the potential of expanding to multiple future research paths.

Figures