APPLICATION OF DATA MINING AND MACHINE LEARNING METHODS TO DEVELOP A DISEASE DIAGNOSIS SYSTEM BASED ON GENE EXPRESSION DATA

Abstract

This monograph addresses the pressing scientific and practical problem of devel- oping and applying methodological foundations of information technology for pro- cessing gene expression data. This technology integrates gene ontology analysis, cluster-bicluster analysis, and deep learning methods for solving tasks in the field of bioinformatics. Its distinctive feature is higher adequacy in disease diagnosis compared to existing methods, achieved through hybridising existing methods and algorithms for big data processing, optimization of model hyperparameters using quantitative quality criteria, and considering the type of data being studied. The relevance of the research topic is underscored by the current absence of adequate in- formation technology for processing gene expression data to identify significant and mutually correlated genes capable of diagnosing the diseases with high confidence and predicting their further development at the genetic level by modelling changes in the expression of target genes and their impact on the studied object. The ef- ficiency of the diagnostic process can be enhanced through model hybridization, which involves the comprehensive application of various methods and algorithms to improve the reliability of decision-making at the corresponding stage. This ap- proach necessitates the development of hybrid quality criteria for evaluating the outcome at each stage. Another way to enhance the effectiveness of gene expression data processing technology is the use of method ensembles, followed by compar- ing the results obtained by each method using appropriate quality criteria and calculating a comprehensive criterion for making the final decision on the model structure. The monograph can be interested for scientists specialized in the fields of both development and applying data science techniques in various fields of scientific research.