Abstarct: The city of Mashhad is the second largest city of Iran and has aboat 3.5 milion population. The existence of various sources of pollutants exacerbates the problem of air pollution and subsequently presenting a solution to this problem in this city is very necessary. In this study, a model baxsed on multiple regressions (linear method) and another model baxsed on artificial neural network (nonlinear method) is presented to predict the concentration of CO, PM10 and O3 according to climatic conditions in city of Mashhad. The results of the linear and nonlinear models are then compared. The meteorological data of this study include wind speed, relative humidity, wind direction, temperature, precipitation and atmospheric pressure were collected from Mashhad Meteorological Station and temperature inversion and traffic data were also included in the model. Air pollution data (concentration of carbon monoxide and particulate matter less than 10 μm) were obtained from the Environmental Pollution Monitoring Center of Mashhad. In this study, air pollution data were obtained in 2016 & 2017 for Sajad air pollution station in Mashhad. Out of the total data 15% were selected for network testing, 15% for network validation, and 70% for network training. Then the correlation coefficient between air pollutants and climatic elements was investigated, that indicates high impact of wind speed (38%) and air temperature (29.2%) on air pollution. The networks were run with one, two, three and four hidden laxyers for each contaminant and the results, which were for 730 days, were presented as tables and graphs. These results confirm that the ability of the neural network model (with RMSE error of 0.137 for CO) is more than linear regression methods (with RMSE error of 0.572 for CO). Also, using more hidden laxyers and more neurons in the neural network laxyers does not necessarily lead to better performance. Predictions for the next 5 days with mean error of 0.97, 0.95 and 0.98 for CO, PM10 and O3 were obtained by time series method, respectively