| Traditional clustering is a hard division, having the qualities of "either this or that" . The fuzzy clustering establish uncertain describe to classification. It express intermediary nature of sample. However, this effectiveness will be affected after introducing concept of membership as follows: (1) the uncertainty of membership function. There is no universally applicable unified pattern on membership. The best definition of membership is that it is poorly targeted and unwanted, which is subjective. (2) quantitative value of membership. That is uncertainty transfer to certainty.Aiming at the problem of membership in data mining, this paper studies uncertainty mechanism of cloud theory, and create a new model based on cloud theory—Random Fuzziness Model (RFM). The model which set up mapping between randomness (accuracy) and fuzziness (membership), improve the reasonable of the determination of membership by randomness depict fuzziness a deeper level. In process of RFM applied to the clustering, first, calculate membership under different accuracy using Random Fuzziness Model. Secondly, get clustering results under different accuracy by clustering Algorithm. Final, get optima solution by validity function of random fuzziness clustering.This paper demonstrate the improved algorithm using both oilfield dataset and standard datatset IRIS. The results showed that new Algorithm can resolve the determination and quanization of membership,and improve veracity and quality of clustering. |
