In this work, two series of regression models were constructed and tested – one comprising models based on the Random Forest algorithm, a machine learning method, and the other based on linear regression. The models were fitted to the data on traffic flow within chosen intersections in the city of Rzeszów and optimized by manipulating the explanatory variables and input parameters. Construction process and optimization efforts have been extensively documented in this article. The performance of both types of models was evaluated in a series of tests, including fitness to the empirical data, residual distribution, prediction of new data, and model training time. Both kinds of models passed the tests favourably, while pointing out some of the advantages and disadvantages of the regression methods used. The results are illustrated on various charts, and the most interesting parts of the program code used are presented.