Delhi air quality index forecasting using statistical and machine learning models

India is one of the largest countries of the world with a high population density. Being a developing country, this nation is facing enormous challenges in regulating its air quality. Population growth, vehicular ownership, urbanization, demand of energy and industrialization process are driving the pollution level to an alarming state in Delhi which is the capital of India. Accurate air quality index (AQI) forecasting of Delhi will assist Government in making strategic decisions so as to avoid catastrophic deterioration in the quality of healthy air. Motivated from this, in this paper, for the first time a systematic study is made to assess the effectiveness of four promising univariate statistical forecasting models and twelve machine learning (ML) models in predicting the AQI of Delhi. Four accuracy measures including “mean absolute scaled error (MASE), symmetric mean absolute percentage error (SMAPE), root means square error (RMSE) and mean absolute error (MAE)” are used to assess the performance of statistical and ML models. Extensive statistical analyses on obtained results using Wilcoxon Signed-Rank test illustrate the superiority of multilayer perceptron (MLP) and autoregressive integrated moving average (ARIMA) models than other considered models in predicting the Delhi AQI.