Forecasting of monthly air quality index and understanding the air pollution in the urban city, India based on machine learning models and cross-validation

In this paper, the study focuses on the forecasting of the Air Quality Index (AQI) using linear regression, random forest, and decision tree regression models in Delhi City. The AQI is a crucial metric for monitoring air quality and provides information on the level of air pollution and its potential health risks. The main research aims to develop forecasting of AQI in three scenarios based on the air pollutants data. Monthly average Nitrogen dioxide (NO₂₎, Sulfur dioxide (SO₂), Oxygen (O₃), and Particle matter (PM_2.5) data from 1987 to 2020 were included. The research was executed in two steps: preprocessing datasets, plotting the datasets, and analyzing them in the first step, and training and testing the model's accuracy in the second step. The datasets were divided into training and testing sets also we forecasted the AQI in three scenarios based on the different input variables. Feature importance was used for the selection of model input variables. Results of the study area compared the Machine Learning (ML) models in three scenarios best performance models such as Decision Tree Regression (DT) (R² = 0.99, RMSE = 0.81), Random Forest (RF) (R² = 0.98, RMSE = 16.64), and RF (R² = 0.99, RMSE = 0.27), respectively. The results of DT and RF models showed high prediction performance compared to other models in the first, second, and third scenarios, respectively. The results of 10-fold cross-validation models are cross-validated to all models, which is the RF model is best other than the models in three scenarios. Hence, the cross-validation of all ML models so important for the selection of the best model for forecasting AQI in Delhi City. The results can be helpful to urban policy makers in the Delhi city.