Regression model and prediction of real estate market development[Regresní model pro predikování vývoje cen na nemovitostním trhu]

The article deals with finding a suitable approach for prediction of real estate prices. A delayed multiple linear regression model appears to be appropriate. This model has been created and has been validated. The model is also suitable for use in routine practice by experts and appraisers, thanks to the simplicity of the calculation without ownership of any computing program. An expert or an appraiser, thanks to the created model, is able to predict the development of the real estate market. The result is bound to the accuracy of the input data. The backward regression in the selection of significant quantities has highlighted the relevance or inconsistency of factors that can enter the prediction of real estate market development. Statistical determinations were performed to verify compliance with the assumptions of regression models, among which the most significant removal of the occurrence of multi-collinearity, the Durbin-Watson test for detecting the presence of autocorrelation of residues, White's test to determine the presence of heteroskedasticity. Removing statistically less significant variables resulted in the number of 6 explanatory variables. The calculations were carried out in the GRETL statistical software and, in the alternative, in MS Excel. The novelty of the overall approach lies in the use of mathematical and statistical methods, which are not used in the usual practice of experts and appraisers.