Abstract: The problem of developing an optimal structure of N-version software system presents a kind of very complex optimization problem. This causes the use of deterministic optimization methods inappropriate for solving the stated problem. In this view, exploiting heuristic strategies looks more rational. In the field of pseudo-Boolean optimization theory, the so called method of varied probabilities (MVP) has been developed to solve problems with a large dimensionality. Some additional modifications of MVP have been made to solve the problem of N-version systems design. Those algorithms take into account the discovered specific features of the objective function. The practical experiments have shown the advantage of using these algorithm modifications because of reducing a search space.

• Publication Date: 11-Oct-2015
• DOI: 10.15224/978-1-63248-064-4-83
• Views: 0
• Downloads: 0

Abstract: A hybrid model (in this paper we call it D-DANP) with the integration of DEMATEL (Decision-Making Trial and Evaluation Laboratory) and DANP (DEMATEL-base ANP) techniques is a popular and effective tool which can output some visual maps to summarize those complex human thinking by its systematization method, and provide some vital information to decision maker. D-DANP model uses questionnaire to collect raw data from expert. Therefore, the most important thing is to collect effective and reliable sample by confirming the quality of the expert. The D-DANP uses the function of consistence examination to verify the reliability of raw data which can influence the outcome of INRM directly and disturb the interpretation of result. However, the D-DANP can not provide the more reliable INRM without finding the outlier form raw data. This paper intends to propose a new modified hybrid model based on the D-DANP model, to develop trustworthy INRM by eliminating outlier which can be found by cluste

• Publication Date: 11-Oct-2015
• DOI: 10.15224/978-1-63248-064-4-84
• Views: 0
• Downloads: 0