The paper suggests a method of representing knowledge in the form of logical equations of a special type. We show that deductive inferences about the salient data features in a knowledge base can be represented by a set of logical equations. We describe how logical equations with finite predicates can be successfully used for the description of logical links between discrete features, and how this can be applied to pattern recognition and data mining. New knowledge about logical links between discrete features in the data can be obtained by eliminating variables from these equations with the help of the operation {E is mirror write}, and we describe this process in detail. We consider the application of the operation {E is mirror write} to a logical equation as an analogue to a query in a database. The process results in a dependence between the features subsequent to application of the elimination procedure that is easier to interpret than the dependence represented in the original equations, and is obtained without the need for exhaustive searching. We recursively define a class of finite predicates that allow eliminating variables without increasing the size of the original equation, and show how this can be applied for knowledge discovery using logical data modelling.
展开▼
