Constraint Satisfaction Method to Search Patterns in the Data Taking into Account the Hierarchy of Features

Abstract

This article considers how the analysis of knowledge from subject domain ontology can help discover the most interesting and well-interpreted patterns in data. For this research area, the term “Pattern discovery guided by ontology” is used in the literature, and ontologies are considered as a means of semantic pruning of the search space. The analysis of ontologies in the pattern discovering can significantly reduce the enumeration of alternatives by pruning the search space, and also allows you to consider the elements of patterns at various levels of abstraction. The proposed approach to Data Mining is based on compact representation of the training sample using specialized matrix-like structures and the application of original inference procedures in these structures. Research lies at the intersection of such areas of artificial intelligence as data mining and semantic technologies for the representation and processing of information.