Özet:

The paper proposes a procedure for the integrated application of methods for scenario analysis and prediction, represented by graphs of the «tree» type. The task on analysis of risks in software projects has been considered, the cause of which are the possible programming errors that lead to failures in the operation of systems and software. The joint use of a failure tree and a probability tree makes it possible to generate the sequences of scripts for the implementation of an adverse event, whose main cause is possible defects or errors in software or data, as well as to assess the probabilities of their realization. Such an approach allows the identification of the overall result of the influence of certain risk-forming factors (defects) on the development of possible negative consequences (failures and malfunctions) or damage to the operation of complex software systems. This makes it possible to timely identify and propose effective mechanisms to manage software risk in order to reduce and eliminate them. A procedure has been proposed for aggregating individual probabilistic expert assessments of the occurrence of a risk event. Such an approach makes it possible to obtain group expert estimates assessing the feasibility of a risk event based on the constructed system of random events into a generalized expert assessment. The probabilities of the occurrence of a risk event, thus obtained, are used when constructing a probability tree and calculating the ratios of probabilistic inference using it. Aggregation of individual expert estimates is carried out by combining them based on a mathematical apparatus of the theory of evidence and the theory of plausible and paradoxical reasoning. It was established that in order to improve quality of the results of combining it is appropriate to establish an order for combining expert evidence and apply one of the rules of conflict redistribution as a combination rule.Numerical calculations of the proposed procedure for integrated application of a failure tree and a probability tree are provided. The results obtained make it possible to run a more in-depth analysis of the examined software systems and objects, and are aimed at improving the quality and effectiveness of managing risks in software projects caused by defects in programs and data Author Biographies Igor Kovalenko, Petro Mohyla Black Sea National University 68 Desantnykiv str., 10, Mykolaiv, Ukraine, 54003 Doctor of Technical Sciences, Professor Department of Software Engineering

Anahtar Kelimeler: