Semantic aspects of data sparsity description in multidimensional information system

Maxim Fomin
Metadata in information systems based on a multidimensional approach can be described through setting the parameters of cells in the multidimensional data cu-be. Metadata can be constructed by connecting classification schemes. Each clas-sification scheme is a hierarchy of members related to a separate structural com-ponent of the observed phenomenon. The method is based on the identification of groups of members that are associated with groups of members of other dimen-sions. Groups of members of different dimensions are used to build clusters of member combinations. Combinations in cluster are formed by the Cartesian product of groups of members. The metadata of the information system is repre-sented as a set of possible member combinations, which is formed as a set of clusters. To solve this problem, the observed phenomenon is considered as a set of structural components. It is necessary to select separate sets of dimensions that are semantically related to the structural components of the observed phenome-non. The semantic connections revealed during the analysis of the structural com-ponent allow us to build a hierarchy of groups of members and represent them in the form of a classification scheme associated with the structural component. In information systems with a large set of aspects of analysis data cubes are charac-terized by a large sparsity. Classification schemes describe the individual aspects of metadata associated with the individual structural components of the observed phenomenon. Combining classification schemes makes it possible to obtain a complete description of metadata by dividing the analytical space into structural components.