A Novel Weighted Fuzzy Clustering Analysis Based on AFS Theory

In the framework of AFS (Axiomatic Fuzzy Sets) theory, We propose A novel weight fuzzy clustering algorithm, which is totally different from the traditional clustering algorithm based approaches. The novel weighted fuzzy clustering algorithm has three main advantages: Firstly, the procedures of the proposed algorithm are more transparent and understandable, and the clustering results not only have definite linguistic interpretation,but also have a weight assigned to each attribute in the cluster description to make the weightpsilas effect on the clustering reflect the importance of the attribute. Secondly, the predefined distance function and objective function are not required, and the cluster number need not be given in advance. Last, the data types of the features can be various data types or sub-preference relations,even human intuition descriptions. To evaluate the performance of the proposed weighted fuzzy clustering algorithm, we consider three well-known benchmark clustering problems-Iris data,Wine data and Wisconsin diagnostic breast cancer data.