Compensation model of multi-attribute decision making and its application to N-version software choice

Abstract

Multi-attribute decision making deals with discrete finite set of alternatives. The solution to the multi-attribute decision making problem is the choice of an alternative from the set of all possible alternatives on the base of usually contradicting attributes. In this paper, a new multi-attribute decision making model is presented. The proposed model develops a linear compensatory process for the interconnected attributes. It concerns the overall ranking of the alternatives based on the attribute-wise ranking as well as the interaction and the combination of the attributes. The compensation model of multi-attribute decision making is applied to N-version software selection. N-version programming is one of the well-known software development approach which ensures high dependability and fault tolerance of software. However, the problem of extra resource involvement arises since the N-version programming stipulates program redundancy. A set of characteristics/attributes have to be considered when choosing an optimal variant of N-version software. The proposed compensation model of multi-attribute decision making provides a solution to this problem. Additionally, a case study on choosing N-version software for a real-life information and control system problem is provided to verify the correctness of our model.