Principle of information diffusion
Fuzzy Sets and Systems
Fuzzy Measure Theory
Techniques of Cluster Algorithms in Data Mining
Data Mining and Knowledge Discovery
Numerical methods for fuzzy clustering
Information Sciences: an International Journal
Hi-index | 0.00 |
Fuzzy comprehensive evaluation cannot reasonably differentiate the close membership values, e.g. 0.70 and 0.69. When the results have to be decided on the basis of maximum fuzzy membership value, some related information among similar objects may be neglected. At the same time, supervised fuzzy clustering analysis selects the threshold according to subjective experience. But different users may give different thresholds, and different thresholds may further get different clustering results. Integrating both fuzzy comprehensive evaluation and fuzzy clustering analysis in a unified way, this paper proposes a fuzzy comprehensive clustering method based on the maximum remainder algorithms and maximum characteristics algorithms. First, the principle of fuzzy comprehensive clustering is given. Based on the membership matrix of fuzzy comprehensive evaluation, fuzzy similar matrix is generated. Then a fuzzy equivalent matrix is produced from the fuzzy similar matrix. According to the fuzzy equivalent matrix, fuzzy clustering is implemented via the maximum remainder algorithms on the basis of fuzzy confidence level. And the grades of the resulting clusters are computed by using the maximum characteristics algorithms. Finally, a case study is given on land grading in Nanning city, the results of which show the proposed fuzzy comprehensive clustering method is able to overcome the disadvantages of either fuzzy comprehensive evaluation or fuzzy clustering analysis.