Algorithms of identification and elimination of mistakes in knowledge bases of expert diagnostic systems

In this work diagnostic systems which on the functions are intended for the help to the user in permission of the following problem situation are considered: it is required to determine by a set of values of the private signs describing a condition of an object of a research. what properties this object has. Such tasks can arise, for example, in medical diagnostics (on a set of the symptoms describing a condition of the patient, the doctor has to take out the conclusion about the nature of a disease); in technical diagnostics (the reason of failure or breakage is determined by a number of indicators of work of a system); in geology (on the basis of data of geological exploration to draw a conclusion on existence of mineral deposits), etc. Unfortunately, for the solution of similar tasks use of methods of machine learning or deep machine learning therefore the only exit is use of expert mechanisms is not always possible. The research objective consists in forming of the knowledge base for the solution of problems of diagnostics which have to comprise decisive rules of the skilled experts allowing to build classification of all possible conditions of an object in previously structured problem area which classes are subsets of the states having the same property. Research methods. Methods of binary classification, the theory of counts are applied to the solution of problems of forming of the consistent knowledge base. When proofreading in classification of states there is no possibility of unambiguous carrying out necessary changes in classification of earlier identified states as elimination of mistakes cannot be carried out separately for each class. In this case to each state only one class of accessory therefore possible addition of a class of accessory to some states results in need of an exception of the accessory classes which are earlier appointed by it that is reached by use of the directed counts has to be appointed.