Development and validation of fuzzy multicriteria decision making models

2017-02-28T03:06:44Z (GMT) by Kuo, Yu-Liang
Fuzzy multicriteria decision making (MCDM) has been widely used in ranking a finite number of decision alternatives characterised by fuzzy assessments with respect to multiple evaluation criteria. The MCDM methods suitable for solving a given decision problem usually differ in their normalisation process and aggregation process for handling the performance ratings of the decision alternatives and the weights of the evaluation criteria. The overall preference of a decision alternative is obtained by aggregating the criteria weights and the performance ratings of the alternatives, on which the ranking is based. Due to their structural differences, these methods often produce inconsistent ranking results for the same fuzzy MCDM problem. To address this issue, this study develops a novel approach for the development and validation of fuzzy MCDM models. The approach incorporates three normalisation methods, three aggregation methods, and a α-cut based defuzzification method to develop fuzzy MCDM models. The α-cut based defuzzification method allows the decision maker’s attitude on fuzzy assessments to be incorporated into the decision making process. To examine the validity of the fuzzy MCDM models available for a given decision problem, a new validation process is developed based on the fuzzy clustering technique to assist in selecting a valid outcome from the inconsistent ranking results produced by these models. To demonstrate the effectiveness of the fuzzy MCDM model development and validation approach, three practical applications under various decision contexts are conducted. The first application is about the airport performance evaluation problem. This study selects 12 Asia-Pacific major international airports as the decision alternatives of the evaluation problem and identifies 19 quantitative and qualitative evaluation criteria under the airport operator, passenger, and airline dimensions. Based on three normalisation methods and two aggregation methods, six fuzzy MCDM models are developed which produce inconsistent ranking results for the evaluation problem. The ranking validity of the six models is examined by the validation process using fuzzy clustering and the most valid model is selected. The second application is concerned with the scrap metal buyer selection problem. This study considers five recycling companies in southern China as the decision alternatives of the buyer selection problem and identifies four qualitative selection criteria under the economic and environmental dimensions. Based on three normalisation methods and three aggregation methods, seven fuzzy MCDM models are developed which produce inconsistent ranking results for the selection problem. The ranking validity of the seven models is examined by the validation process using fuzzy clustering and the most valid model is selected. The third application deals with the non-ferrous scrap metal supplier selection problem. This study considers 15 scrap metal suppliers as the decision alternatives of the supplier selection problem and identifies five quantitative and qualitative selection criteria for a non-ferrous scrap metal buyer. Based on three normalisation methods and three aggregation methods, seven fuzzy MCDM models are developed which produce inconsistent ranking results for the selection problem. The ranking validity of the seven models is examined by the validation process using fuzzy clustering and the most valid model is selected. With the development of the approach and the three empirical applications, this study makes significant methodological and practical contributions. The approach addresses the validity issue of the cardinal rankings generated by different fuzzy MCDM models. In practical applications, the subjective attitude of the decision maker is effectively incorporated into the decision making process. With its simplicity in both concept and computation, the approach has a general applicability for solving general MCDM problems, and is particularly suited to decision situations where the ranking results produced by different fuzzy MCDM models differ significantly.