Air carbon monoxide forecasting using an artificial neural network in comparison with multiple regression

This study aimed to forecast air concentrations of CO in Tehran. This research is an analytical-practical method using daily data from Tehran's air quality monitoring stations, weather parameters, time parameters such as 1-day time delay and traffic parameters, the prediction model of air pollution caused by Tehran metropolitan transport. To apply in the decision support system, it was investigated. In the multi-criteria decision-making process, the importance of evaluation indicators is usually taken into account. Gray relationship analysis was used to rank the influencing parameters in air pollution. After estimating the effective parameters, an artificial neural network model was used to forecast the CO concentration using MATLAB software. In the end, the results of an artificial neural network model were compared with the linear regression model. Correlation coefficient and mean square error of the neural network model R = 0.72 and RMSE = 0.69 with linear regression model of R = 0.10 and RMSE = 11.747 were compared. The results of this study showed that the neural network model error is less than the linear regression model. Based on the results of sensitivity analysis, hot/cold season parameters, 1-day delay, 2-day delay, day of year, and month of the year have the most effect on the concentration of CO in Tehran.