Air Quality Monitoring and Forecasting System using IoT and Machine Learning Techniques

Air pollution has been a growing concern in the twenty-first century, affecting the surrounding environment and public health. The previous studies have recently undertaken significant research on air pollution and air quality monitoring. Unfortunately, this area continues to be challenged by unresolved issues. This paper proposes an IoT-based Air Quality Monitoring and Forecasting System to monitor and predict air pollution for a specific area based on various pollution factors. Using Arduino UNO R3 and various low-cost sensors, our IoT system can collect and monitor pollutants, such as PM2.5, CO2, CO, as well as temperature and humidity. The air quality data was collected for several months. To overcome the problems of instability of low-cost devices in monitoring, machine learning (ML) algorithms, such as K-Nearest-Neighbour (KNN), Expectation-Maximization (EM), Multiple Imputation by Chained Equations (MICE), and Autoregressive-Moving-Average (ARMA), are applied to address missing data and outliers due to technical issues. The KNN model outperformed all others in terms of RMSE, MSE, MAE, R-squared, and execution time. Then, Autoregressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) algorithms are applied to predict future air quality. The result shows that our system can predict the air quality factors over the next hour with the highest accuracy at 96 %. Finally, a web interface was created to monitor and forecast air quality in real-time.