Air Quality Prediction in Smart Cities Using Cloud Machine Learning

In the last couple of years, significant advances have been made in implementing Internet of Things frameworks in the architecture of smart cities to enhance people’s living standards. However, poor, smart city planning might pose specific concerns regarding energy usage, infrastructure, pollution, security, and privacy. In this paper, we will address one such challenge: air pollution, which can impact the environment, humans, and other living species. The air quality index is calculated from the pollutants particulate matter, ozone, nitric oxide, nitric x-oxide, carbon monoxide, benzene, toluene, sulfur dioxide, xylene, ammonia, and nitrogen dioxide to address air pollution. We utilize machine learning techniques to forecast the air quality index. To predict air quality, we implement regression models such as decision tree regression and random forest regression. Additionally, we employ cloud computing to minimize execution time. For this, we execute the same machine-learning models on Amazon SageMaker’s Jupyter Notebook and record their execution times. For the Indian smart cities dataset, these models are compared using evaluation metrics and execution time. The results show that cloud computing reduces execution time for all models while maintaining accuracy. The random forest algorithm achieves better accuracy than the other two models.