Reliability of LSP Criteria

摘要

We analyze the reliability of results obtained using the Logic Scoring of Preference (LSP) method for evaluation and comparison of complex systems. For each pair of competitive systems our goal is to compute the level of confidence in system ranking. The confidence is defined as the probability that the system ranking remains unchanged regardless of the criterion function parameter errors. We propose a simulation technique for the analysis of the reliability of ranking. The simulator is based on specific models for selection of random weights and random degrees of andness/orness. The proposed method is illustrated by a real life case study that investigates the reliability of evaluation and selection of a mainframe computer system.