This work is devoted to the development of technology for analyzing heterogeneous data based on the concept of multidimensional representation by including multimodal data in the form of granular structures in the fact-measurement model. Modern methods of information support for decision making are undergoing radical transformation in the context of digitalization. In addition to the classic tasks of minimizing uncertainties and increasing the accuracy and reliability of descriptions of objects and processes, there is now a need to quickly collect and analyze information from diverse sources: relational and post-relational databases, intersystem interaction formats, and unstructured data (text, images, audio, and video). Granular computing methods make it possible to extract structured data and knowledge needed for decision making from unstructured fragments of information. Typically, information granules are used to solve strictly algorithmic problems, while the extracted data can be included in the general information model of the studied subject area for deeper analysis and obtaining new knowledge in the research mode. The main idea of the work was to develop a classical fact-dimension model by interpreting the concepts of information granule and granular structure in the concept of multidimensional data representation as new elements of a heterogeneous multidimensional model. This article compares the concepts of analytical and computational granularity, proposes an approach to integrating information granules into a multidimensional data model, and examines an example of analyzing the results of an experiment on the formation of a carbon film on a sapphire substrate, where the experimental parameters are presented in relational form and the characterization parameters are obtained based on the analysis of atomic force microscope images.
Anna Vladimirovna Korobko (Mon,) studied this question.