Characterization and problem analysis of regional economical development is necessary to identify and put into effect such economical development policy that would stimulate the increase of quality of life and balance the development of territory in the whole country. In the analysis a significant role is also played by collection of statistical data and application of data analysis methods, including also the linear regression which, by using the available statistical data, allows obtaining adequate polynomial regression models for prediction of yet unknown future data values. Typically, the standard regression modelling methods, implemented in popular statistical software packages, require user to predefine a number of parameters, e.g., the maximal order of the model (for generating the full set of basis functions from which to select the subset) or the significance threshold with which the basis functions are included in the model. However, in many cases the necessary values of the parameters are unknown and need to be guessed, resulting in a non-trivial (and long) trial and error process. We consider a more convenient and potentially also a more efficient approach than the subset selection – letting the regression model induction method itself construct the basis functions necessary for creating the model without the requirement to predefine them like in subset selection. The method also does not require setting the significance threshold.