A.A. Kuzmin, M.N. Kuzmina, S.F. Jatsun, A.A. Naser

Knowledge base is the basis of medical decision support systems. Domain is described in this database in the form of logical expressions. Often fuzzy logic is used to improve the accuracy of a mathematical framework. There are several approaches to solve practical problems using fuzzy logic. L. Zadeh suggested that one approach is based on the theory of fuzzy sets, which are represented by membership functions to the classes of states under investigation. Another approach uses the theory of E.Shortliff-s certainty factors or trust network. In this approach, there is speculation that the two mutually reinforcing evidence must strengthen confidence in the conclusion (prediction, diagnosis). Pointing in one direction several evidences can not be fully offset by evidence pointing in the opposite direction. As an illustration of the theory of certainty factors, the article shows how to change confidence in the diagnosis of "hypertension" in the registration of indirect evidence, "edema" and " thirst." It is recognized that the edema and thirst can be signs of kidney damage, and kidney are target organs for hypertension and they may be affected by prolonged pressure in the arterial line. Moreover, conditionally accepted, that trust in the symptom "edema" more reliable in terms of setting the output of the kidney disease, compared with the symptom of "thirst" which you may receive from the larger reasons. And besides, even if the kidney damage is present, then it can not be unambiguous criterion that a patient has high blood pressure, because the kidneys may be affected not only due to high pressure. As an illustration of the combined approach, which combined the classical methods of fuzzy decision rules for L. Zadeh, as well as methods for the synthesis of fuzzy rules proposed by E. Shortliff, the method of aggregating data on risk factors for hypertension is described. If the theory of certainty factors is the aggregation of confidence factors in the hypothesis, the proposed method is the aggregation of the values of membership functions, each of which can be regarded as a special confidence factor in the hypothesis. And the more each registered parameter out of the norm corridor, the greater its contribution to aggregate value and the greater the overall risk of complications of the disease. The proposed approach simplifies the computational procedures and decision rules in comparison with traditional methods.