Air Quality Prediction in Mumbai city using Machine Learning-based Predictive Models
Rapid urbanization and industrialization leads to major environment problem of air pollution. Air Quality (AQ) essentially must be constantly supervised, assessed and forecasted to assure healthier conditions to live for human, animals and vegetation life. U. S. Environment Protection Agency (EPA) defines Air Quality Index (AQI) which requires accurate and precise sensor readings. High level of Particulate Matter 2.5 (PM2.5) has been considered to be very hazardous among all pollutants present in the air, making its level to be continuously monitored, predicated and controlled. The AQ becomes major problem in Mumbai City, India and State Government-Municipal Corporation is taking efforts for policy reforms. In this paper various machine learning approaches has been analyzed as it provides better results for classification and predication for AQ. The aim of this paper is to compare different machine learning and deep learning models like Autoregression (AR), Deep Neural Network (DNN), Recurrent Neural Network, Long Short Term Memory (LSTM) and Bidirectional LSTM for prediction of pm2.5 pollutant with time lag of 1, 4, 8, 12 and 24 hours. The RMSE and R2 value are taken as performance metrics for evaluation of models. The simulation results show that bidirectional LSTM outperformed over RNN and LSTM with RMSE 19.54 and R2 value of 0.66.