Air pollution prediction using semi-experimental regression model and Adaptive Neuro-Fuzzy Inference System

journal article in Web of Science

en

Lifestyle development and increasing urbanisation and consumption of fossil fuels, monitoring and controlling air pollution have become more important. This study has used the available data of key pollutants to predict their future status through time-series modelling. Most researchers have employed Autoregressive Integrated Moving Average and Logistic Regression techniques, and Adaptive Neuro-Fuzzy Inference System has rarely been used to analyse time-series data. Traditional time-series forecasting models assume a linear relationship between variables, while there are nonlinear and complex components in air pollution modelling. This study aimed to respond to this limitation by improving the accuracy of the daily prediction of pollutants via time-series data analysis by using Adaptive Neuro-Fuzzy Inference System modelling. A nonlinear multivariate regression model was developed and experimentally refined to obtain the least error possible. Data on pollutants containing CO, SO2, O-3, and NO2 are collected from a single monitoring point in Tehran. The process of the developing the model begins by breaking down the data sets into training, testing, and validation set at a random ratio of 80%, 10%, and 10%. For the prediction of CO, SO2, O-3, and NO2, the coefficients of determination are calculated as 0.8686, 0.8011, 0.8350 and 0.7640, and these values for the semi-experimental model were 0.8445, 0.8001, 0.7830 and 0.7602. According to the performance indicators of both models, Adaptive Neuro-Fuzzy Inference System is more accurate in predicting time-series data than regression models. Reliable forecasting of future air quality would help governments develop policies and regulations to protect humans and ecosystems and achieve sustainable development. (C) 2020 Elsevier Ltd. All rights reserved.

Lifestyle development and increasing urbanisation and consumption of fossil fuels, monitoring and controlling air pollution have become more important. This study has used the available data of key pollutants to predict their future status through time-series modelling. Most researchers have employed Autoregressive Integrated Moving Average and Logistic Regression techniques, and Adaptive Neuro-Fuzzy Inference System has rarely been used to analyse time-series data. Traditional time-series forecasting models assume a linear relationship between variables, while there are nonlinear and complex components in air pollution modelling. This study aimed to respond to this limitation by improving the accuracy of the daily prediction of pollutants via time-series data analysis by using Adaptive Neuro-Fuzzy Inference System modelling. A nonlinear multivariate regression model was developed and experimentally refined to obtain the least error possible. Data on pollutants containing CO, SO2, O-3, and NO2 are collected from a single monitoring point in Tehran. The process of the developing the model begins by breaking down the data sets into training, testing, and validation set at a random ratio of 80%, 10%, and 10%. For the prediction of CO, SO2, O-3, and NO2, the coefficients of determination are calculated as 0.8686, 0.8011, 0.8350 and 0.7640, and these values for the semi-experimental model were 0.8445, 0.8001, 0.7830 and 0.7602. According to the performance indicators of both models, Adaptive Neuro-Fuzzy Inference System is more accurate in predicting time-series data than regression models. Reliable forecasting of future air quality would help governments develop policies and regulations to protect humans and ecosystems and achieve sustainable development. (C) 2020 Elsevier Ltd. All rights reserved.

Air pollution prediction; Adaptive neuro-fuzzy inference system; Semi-experimental model; Nonlinear regression; Time-series data; Sustainable development; NETWORK; ANFIS; OPTIMIZATION; FRAMEWORK

10.06.2020

Elsevier Ltd

ELSEVIER SCI LTD, THE BOULEVARD, LANGFORD LANE, KIDLINGTON, OXFORD OX5 1GB, OXON, ENGLAND

0959-6526

261

121218–121218

16