An approach is proposed to the solution of formalized problems of assessment of the activity that produces and maintains software systems (SSs). Such assessment is realized by using expertises that form a new assessment process adequate to the activity needs and specifics with an environment common to the expertises. The following mathematical apparatus is elaborated for expertises: a framework (target functions and executing mechanisms), a model and methods (formalisms for improving the quality and reusing the results of expertises) of an assessment process, and tools for integrating the apparatus into software development management processes. The approach is theoretically justified. Prospects of developing the proposed approach are described.Keywords: production of software systems, assessment problem, mathematical apparatus, expertise technology, model of an expert assessment process, ontology, homomorphism, metrized similarity, value tree, validity of an expert decision.
INTRODUCTIONThe professional activity in creating software systems (SSs) assumes industrial features and status. It is based on technological standards, regulations, and estimates of the results of actions directed towards the design of SSs on the so-called Basic Process (BP) (involving analysts, programmers, and testers) and also towards the management over BP using the so-called Management Process (MP) (involving customers, managers, and appraisers).Taking into account the specificity of the BP [1, 3-6] and MP [2][3][4][5], the activity in producing SSs, i.e., the Production Activity (PA), can be represented by the following cortege:where DA P is the object domain of making decisions on management over P , RQ P are the requirements of a process P on the process complementing it in PA cortege (1), DC , DI , DE, and DM are the subdomains of the object domain of software engineering (ObSE) that correspond to its disciplines singled out in [6][7][8], namely, Economics, Industry, Engineering, and Management. The achievement of the main objective of activities in PA, namely, the improvement of the quality of SSs, presumes [1-5] repeated assessments of processes P of activity (1) in their environment, i.e., the fix of current and predicted values of efficiency criteria (for managed or monitored objects P ), achievement levels (for business objectives of P ), and criteria of prospects (for actions directed towards achieving objectives during the execution of P ).For the criteria of efficiency of the process BP, the following groups of methods [1-3] were formed: -collective decision-making methods such as Delphi (O. Halmer and N. Dalkey), brain storm (A. Osborn), and their modifications (M. Turoff and S. Faure), which are universal but excessively resource-intensive;
6381060-0396/09/4504-0638 -assessment of the quality and laboriousness of an SS project on the basis of Bayesian and neural networks that require the adaptation of a network to the project but do not provide the aggregation of views of participants with different roles;-special expert-...