DETERMINING THE NUMBER OF CRITERIA FOR EXPERT RATING OF ANALYSIS OBJECTS OF INFORMATION TECHNOLOGY PROFILE

The aim of this work is to determine the functional relationship between the quality of the rating procedure for difficult-to-formalize objects of analysis (DFOAs), which include innovative information technology proposals, and the number of evaluation criteria used. The authors consider the content of these criteria as a necessary condition for the effectiveness of the DFOA rating procedure, while their formalized and justified quantity constitutes a sufficient condition for its efficient execution. The objectives of the study include modelling the probabilistic reliability of the criteria-expert evaluation (CEE) procedure for the predictive rating of DFOAs and identifying the interrelationships between its identification parameters. Furthermore, the authors consider this algorithmized CEE procedure for various DFOAs, which include the predictive quality of innovations, a significant component of modern information technologies, where the functionally creative capabilities of neural networks are increasingly used. This is the main perspective for developing the qualimetric approach to DFOA rating proposed in the article. The research methods are based on the apparatus of applied qualimetry of DFOAs and the theory of complex system reliability. This enabled the analytical derivation of a relationship between the set of conditional covariances of individual quantitative values of the evaluation criteria and the coefficients of their influence on the final DFOA rating, such as the predicted level of potential effectiveness of a specific innovative technological proposal. The key new result of the research is determining the functional dependence of the probabilistic reliability of the rating procedure on the information content of the evaluation criteria used and their total number. The resulting general probabilistic relationship made it possible to link the predicted reliability of the CEE procedure with the informational content of the applied evaluation criteria and their total number into a unified functional-probabilistic complex. Using the Pareto principle, this relationship is transformed into a calculation dependency, according to which it is realistic to determine a specific number of evaluation criteria depending only on the requirement for their reliability, namely the probability of an adequate representation of the corresponding quality characteristic of the rated DFOA. The paper provides typical examples for calculating the sufficient number of evaluation criteria at various levels of requirements for their content adequacy. This allows for quantitative formalization of the DFOA rating structure, which includes various innovations of information technology content.

rating, expertise, evaluation, number of criteria, probabilistic model, functional relationship, reliability