A Machine Learning-Based Platform for Monitoring and Prediction of HazardousGases in Rural and Remote Areas

Abstract

Air pollution and the release of hazardous and radioactive gases pose serious threats to the environment and human health, often leading to environmental disasters and premature deaths. These risks are especially severe in remote and rural areas, where access to healthcare, emergency services, and continuous monitoring is limited. In such regions, the absence of real-time monitoring systems makes it difficult to detect indoor gas-related incidents early, causing authorities and healthcare providers to respond only after serious harm has occurred. To address this challenge, this paper presents a digital decision support system that combines Internet of Things (IoT) technology with Machine Learning (ML) to monitor and predict indoor hazardous gas incidents. The system is built on the Rural THINGS IoT platform, developed by the University of Beira Interior, Portugal, which continuously monitors air quality in rural and remote environments. By collecting real-time sensor data, the platform helps assess environmental conditions and warns residents and stakeholders about potential exposure risks. The proposed system uses an advanced ML architecture that combines bidirectional and unidirectional Long Short-Term Memory (LSTM) layers to analyze time-series data and predict future gas concentration trends. Validation using a real testbed demonstrated that the model effectively predicts hazardous gas patterns, enabling timely alerts and preventive actions. This approach supports improved public safety, proactive healthcare responses, and better environmental risk management in underserved regions.