Abstarct: Traffic flow forecasting has attracted a lot in the last few decades. This is still a challenging issue for traffic and transportation researchers. Due to the random characteristics of the traffic flow, accurately predicting traffic will not be easy. Therefore, to solve this problem, many techniques have been used to model and predict traffic flow. In the present study, the parametric method, ARIMA (AutoRegressive Integrated Moving Average), non-parametric method, artificial neural network and deep learning method, LSTM (Long short-term memory) to predict the volume of road traffic It has been used for 6 round trip routes leading to Mashhad and the results of these three methods have been compared with each other. The results show that the LSTM deep learning method performs better than the other two methods. The values of RMSE, MAPE and AADT indices for these three methods are significantly different from each other. The difference between RMSE, MAPE and AADT indices of ARIMA model and LSTM deep learning model is 0.72, 24% and 23%, respectively Also, the difference between the values of these indicators for the artificial neural network model and LSTM deep learning is 0.18, 11% and 6%, respectively, which shows the high ability of deep learning methods compared to parametric and non-parametric methods.