Open AccessFormation of a knowledge base to support the process of architectural design of software systems
Author(s) -
Gleb Guskov,
A.M. Namestnikov,
Anton Romanov,
Aleksey Filippov
Publication year - 2021
Publication title -
ontologiâ proektirovaniâ
Language(s) - English
Resource type - Journals
eISSN - 2313-1039
pISSN - 2223-9537
DOI - 10.18287/2223-9537-2021-11-2-154-169
Subject(s) - computer science , artifact (error) , software engineering , knowledge base , completeness (order theory) , class diagram , process (computing) , software , engineering drawing , unified modeling language , programming language , artificial intelligence , engineering , mathematical analysis , mathematics
This article describes an approach to knowledge base (KB) formation for automating the process of architectural de-sign of software systems (SS) based on the experience of previous projects. Software architecting is the presentation of software systems in the form of design artifacts and their architecture. When developing a new SS it is possible to im-prove its quality based on the experience of previous projects. The experience of previous projects is successful archi-tectural solutions contained in the knowledge base of the design organization. Such a KB should be formed in the pro-cess of analyzing design artifacts extracted from previous projects: source code, project diagrams, data models, struc-tured text resources, etc. This article describes a KB model of a design organization and a model of the 1C: Enterprise 8 (1C) application solution as an example of a design artifact. The article also presents a method for generating frag-ments of a KB in the process of analyzing an applied solution for the 1C application and a method for generating use-case diagrams based on the KB content. A set of experiments was executed to evaluate the adequacy of the proposed models and methods. The results of experiments for assessing quality in terms of accuracy (the presence of elements of the expert diagram in the generated diagram) and completeness (the presence of elements in the generated diagram that are absent in the expert diagram) are presented. According to the results of the experiments, the average value of accuracy is 0.875, and the completeness is 0.6.