This paper is dedicated to comparative analysis of parametric and semiparametric group regression models through the consideration of forecasting of the inland railway passenger conveyances of regions of Latvia. The suggested models contain such factors for each region of Latvia as: number of enterprises per a unit of territory, density of the unemployed population, number of buses per a unit of territory, density of population, number of railway station and so on. The corresponding data and the inland railway passenger conveyances are available for considered regions of Latvia for fixed years. As observation we took an inland railway passenger conveyance for a region of Latvia for concrete year. All experimental calculations were performed on the basis of the statistical data taken from Annual Report of State Joint-Stock Company “Latvijas dzelzceļš” and Statistical Yearbook of Latvia 2003. Various tests for hypothesis of explanatory variables insignificance and model correctness in case of data smoothing have been performed. The cross-validation approach for analysis of models behaviour in case of conveyances forecasting has been applied as well. The comparative analysis of suggested models has shown obvious preference of the semiparametric approach.