For example,Бобцов

SOFTWARE TOOLS FOR COMPUTING EXPERIMENT AIMED AT MULTIVARIATE ANALYSIS IMPLEMENTATION

Annotation

A concept for organization and planning of computational experiment aimed at implementation of multivariate analysis of complex multifactor models is proposed. It is based on the generation of calculations tree. The logical and structural schemes of the tree are given and software tools, as well, for the automation of work with it: calculation generation, carrying out calculations and analysis of the obtained results. Computer modeling systems and such special-purpose systems as RACS and PRADIS do not solve the problems connected with effective carrying out of computational experiment, consisting of its organization, planning, execution and analysis of the results. Calculation data storage for computational experiment organization is proposed in the form of input and output data tree. Each tree node has a reference to the calculation of model step performed earlier. The storage of calculations tree is realized in a specially organized directory structure. A software tool is proposed for creating and modifying design scheme that stores the structure of one branch of the calculation tree with the view of effective planning of multivariate calculations. A set of special-purpose software tools gives the possibility for the quick generation and modification of the tree, addition of calculations with step-by-step change in the model factors. To perform calculations, software environment in the form of a graphical user interface for creating and modifying calculation script has been developed. This environment makes it possible to traverse calculation tree in a certain order and to perform serial and parallel initiation of computational modules. To analyze the results, software tool has been developed, operating on the base of the tag tree. It is a special tree that stores input and output data of the calculations in the set of changes form of appropriate model factors. The tool enables to select the factors and responses of the model at various steps and to form easyto-read tables and graphs of functions for the selected parameters. The proposed solution has been tested in the process of verification for “Prognoz_R” software, greatly simplifying the preparation and carrying out of the design calculations for analysis of the developed computational model. Thus, the proposed software tools, due to the lack of universal ready-made solutions, can serve as an effective replacement for manual carrying out of computational experiments. 

Keywords

Articles in current issue