ABSTRACT
Therefore, it is possible to improve the ARPA function by modeling the knowledge and experience
(2011) showed FCM in the context of local image processing and pattern recognition. Chatzis and Varvarigou (2008) developed a new FCM method based on Markov random chains to improve image characteristics extraction. Aside from its application in the world of graphics, FCM has also been applied in control engineering. Nanda et al. (2012) demonstrated the suitability of FCM for machinery noise feature extraction. It can be inferred from the above references that FCM is an efficient, flexible, universally applicable tool in classification and modeling knowledge and experience, especially in the area of graphics. The major limitation of FCM is that the classification is usually established without considering any input on the significance and discreteness of the different types of information or evidence. A number of researchers have focused on improving FCM; Li and Yu (2009) proposed a comprehensive method to determine optimal classification numbers and fuzzy coefficients using a distribution analysis of the data set. The significance and discreteness are usually described by the Shannon entropy, such as in the case of the decision tree algorithm ID3, C4.5. Ma & Lebacque (2013) used the cross Shannon entropy to state the confusion and discreteness. To improve on the standard FCM method, to make an evaluation of the source information possible, it is appropriate to use the Shannon entropy.
