Formation of mathematical models of the preliminary diagnosis of liver diseases based on the methods of the regression analysis

The paper is devoted to creation of the formalized models for liver diseases (hepatitis and cirrhosis), describing interrelation of the resultant diagnosis from the results of inspection of patients. Basic data (in number of 135 indicators) have specific features. The part of the data relying on individual feelings of the patients has subjective character. Besides, there is no considerable part of the data on the following reasons: the data were gathered during decades and among the analyzed indicators there are some, which have not been included before in the list of the surveyed characteristics of the patients; some data on the patients are absent for the unknown reasons. The work presents 24 new indicators, on which the data set is limited by the data on the patients of the last years. In these conditions, it was expedient to construct a et of the formalized models based on the various sets of entrance indicators. On the basis of the program EViews system, all the possible formalized models, from which a set of 33 models with acceptable values of coefficients of determination and criteria of the importance was created. A new criterion of the assessment of importance of the models considering the volume of the basic data, used at its construction. The comparative analysis of the received models is carried out. As a result, in particular, it is revealed that introduction of new indicators allowed to increase significantly quantity of models as 45% of models include new indicators.