Predictive Modeling and Machine Learning Approaches for Real-Time Air Pollution Monitoring in Dhaka City

Abstract

In recent years, Dhaka has experienced rapid urbanization, which has intensified sources of air contamination such as vehicular emissions, industrial activity, and unregulated construction. These factors have contributed to a noticeable deterioration in air quality. For this causes peoplefaces health problems such as asthma, lung disease, and heart disease, especially amongchildren and old people. In this work, I tried to create a device that could observe and alsopredict air quality in live. For collecting data, I used IoT-based sensors that measure PM2.5, PM10, CO and some other pollutants. After collecting the data, I experimented with several traditional and deep learning models, including regression-based, probabilistic, distance-based, and recurrent neural network approaches. These models are used for predicting air qualitylevels and to see which model gives better results. From the comparison, I found that using amix of models can predict more accurately than using only one model. The system can also givean early warning when pollution is very high. I believe this research will be helpful for Dhakacity planners and authorities so that they can take actions to reduce pollution.