IoT-Driven Environmental Monitoring and Healthcare System: Shipbreaking Industry Perspective

Abstract

Bangladesh's coastal region is ecologically significant and has a rich biodiversity. The country is well-known in the world for its shipbreaking industry. However, its commercial operations have severely impacted the coastal habitat. For example, the workers are exposed to a savage environment of effect that makes them vulnerable to a variety of health issues. Thus, the main objective of this research is to monitor the environmental pollution as well as the health condition of the workers of the shipbreaking industry in Sitakunda region of Chattogram in Bangladesh. We introduce an IoT-driven outdoor environmental monitoring system aimed that collects real-time data on five air parameters: PM2.5, dew point, temperature, surface pressure, and humidity, and six water parameters: pH, turbidity, conductivity, temperature, TDS (Total Dissolved Solids), and DO (Dissolved Oxygen) from the Sitakunda Shipbreaking Industry. The sensing devices are connected to a microcontroller board to gather sensor data and transmit the collected data to a cloud server for further processing and analysis. Moreover, we have collected past environmental data from various sources, such as NASA Power View, Dhaka US Console, Ministry of Environment, Forest and Climate Change, and WHO. We have also conducted a patient survey in six different hospitals of Chattogram to evaluate workers' health status, which is a distinctive and challenging aspect of our work. Subsequently, we have analysed environmental pollution using these datasets and health surveys, identifying three types of risk factors such as High, Medium, and Low, for the workers' health conditions in a shipbreaking industry. Finally, we have constructed a machine learning (ML) model using a variety of ML Algorithms, namely Ensemble Method, Extra Tree Classifier, Random Forest, Decision Tree, and KNN that achieved higher accuracy of 81.78% and 92.26% for air and water, respectively. It is observed that the Random Forest Classifier demonstrates the best precision compared to other models based on dataset accuracy. During the summer season, our system identifies a category of "Medium Risk Factor" based on the environmental condition. Our implementation provides a method to determine the affected rate of the environment and the margin for a healthy environment. Consequently, this approach will assist in estimating and informing the public about the hazardous situation of shipbreaking industries, particularly those in the Sitakunda region.