Mass appraisal of multiparameter objects with range dependent variable

Statistical mass appraisal of objects is carried out by constructing various regression models and using iterative procedures to find the coefficients of these models. Practical application of such models can be fraught with certain difficulties (first of all, due to weak noise immunity of these models). In this paper, it is proposed in mass appraisal to move from a continuous dependent variable of regression model to a discrete one, which does not require finding the coefficients of the model and increases its noise immunity. Instead of iterative procedures, any classification algorithm can be used in this case. In comparison with the regression model, classification is a more convenient tool in practical applications, because various information processing mechanisms implemented in appropriate algorithms can be used. In connection with the unsatisfactory quality of the cadastral valuation of real estate conducted earlier in the Russian Federation, a course has been chosen by the Ministry of Economic Development to form and evaluate so called price zones. The value of the price range within the zone depends on the properties of the objects entering the zone. In this regard, the choice a discrete dependent variable in models can be practically used in the formation of price zones. In paper, we offer a simple classification algorithm for the formation of price zones. Application of the algorithm is shown by the practical example of a classification of apartments in various districts of Sochi.