EVOLUTIONARY METHOD OF FACTOR ANALYSIS OF DATA PRESENTED IN THE FORM OF TRANSACTION DATABASES.

The solution of the problem of factor analysis automation in the diagnosis and recognition of images is considered in the paper, and some results of our research in this area are given. The main purpose of the study is to develop an evolutionary method of factor analysis to find hidden dependencies in transactional databases. The use of modern methods of evolutionary search allows forming the groups of similar features. The issues of extracting factor groups from the specified transactional databases are considered in the paper for identifying new knowledge when solving the problems of diagnosis and recognition of images. The proposed method allows extracting the groups of qualitatively similar features from transactional databases. We propose to use the association rules to assess the equivalence of features terms that allows assessing the closeness of relationship between various features, making no demands to the input data and performing the factor analysis in transactional databases. The research results can be used by researchers dealing with the study and analysis of complex objects, processes and systems with the purpose to identify new knowledge, as well as in decision support systems for technical and medical diagnostics. [ABSTRACT FROM AUTHOR]