In the thesis we have established a methodology for the analysis of rhythmicity in count data. We have implemented a set of functions, that we can use directly for the analysis of circadian count data. To analyse this type of data, we need to combine methods for rhythmicity detection with methods for analysing count data. For the purpose of rhythmicity detection and transformation of input data, we have used the cosinor method. The implemented computational method allows to identify the number of components automatically. For the analysis of the count data and the solution of the regression problem we have used five computational models -- Poisson model, generalized Poisson model, zero-inflated Poisson model, negative binomial model, and zero-inflated negative binomial model. The established method allows us to compare and find the most suitable model with the optimal number of components. The method also includes functions to compare the rhythm in dependence of different factors. The complete method was tested on two traffic datasets obtained from the Ministry of Infrastructure of Republic of Slovenia.