Many real-world decision problems require sorting alternatives into ordered classes, and often they involve multiple measures, making them multi-criteria sorting problems. Previous research on applying TOPSIS (The Technique for Order of Preference by Similarity to Ideal Solution) to these practical problems has focused on proposing criteria weights and computing relative closeness, obtained by comparing distances of alternatives to the positive and negative ideal solutions. However, the issue of how to determine the cutoff values has not been attacked before. We propose a general approach to determine optimized cutoff values, with objective weights for the TOPSIS sorting process. These cutoff values are obtained by minimizing the sum of deviations for randomly selected representative alternatives of neighboring classes. The procedure is demonstrated using two public datasets. It is then analyzed and compared with previous research and traditional data mining techniques, and the results demonstrate that TOPSIS is an effective tool for ordered sorting.
關聯:
International Journal of Information and Management Sciences (IJIMS) 35(2), p.175-190
