Prediction of air quality in Tehran by developing the nonlinear ensemble model

By Aydin Shishegaran, M. Saeedi, Anikender Kumar, H. Ghiasinejad. 2020

71 citations recorded by the source corpus.

Abstract Prediction Air Quality Index (AQI) is a useful technique to improve public awareness about air quality in next days that is a great concern in developed and developing countries. In this study, four prediction models are utilized for predicting daily AQI. These prediction models include Auto Regressive Integrate Moving Average (ARIMA) as a time series model, Principal Component Regression (PCR) as a hybrid regression model, combination of ARIMA and PCR as the first ensemble model and, the combination of ARIMA and Gene Expression Programming (GEP) as the second ensemble model. Observed AQI during the years 2012–2015 was utilized to train models, which is named the calibration process. Based on the calibration of each model, four equations were obtained to predict daily AQI for each season separately, and then these equations were used to predict daily AQI for each season in 2016. The maximum negative and positive errors, Mean Absolute Percentage Error (MAPE), and statistical parameters, including the coefficient of determination, root mean square error (RMSE), normalized square error (NMSE), and fractional bias, were utilized to evaluate and compare models. Based on these evaluations, the two best models are specified, and then a novel statistical table, which can specify the distribution percentage of errors, was used to specify the best model for predicting daily AQI in each season. According to the results, model 4, which is the nonlinear ensemble model, is considered as the best model for predicting AQI in all seasons. The results show that the coefficient of determination of model 4 is close to 1, and the values of its NMSE are between 0.012 and 0.51, and the values of RMSE are between 2.870 and 8.125.

Subfields: Automotive Engineering, Environmental Engineering, Health, Toxicology and Mutagenesis

View Paper: https://www.semanticscholar.org/paper/bbd3f0d9b5bc81b67aebe80301098306d1f81535