Research On Intuitionistic Fuzzy Clustering Method : Based On Knowledge Measure Theory

Clustering algorithm is an effective method to classify complex samples,which is widely used in pattern recognition,machine learning and other fields.Intuitionistic Fuzzy Clustering Method(IFCM)is introduced to the intuitionistic fuzzy environment to make it more clearly describe the uncertain information in the objective world.In recent years,IFCM algorithm has been the research focus in the field of clustering.However,the current academic circle has not formed a complete and clear definition of IFCM algorithm,and the algorithm still has some shortcomings,such as the lack of difference in the contribution of different data characteristics to the clustering center and the sensitivity of the algorithm to the initial clustering center.Intuitionistic fuzzy entropy is a measure to measure the uncertainty and fuzziness of intuitionistic fuzzy sets.Many scholars have introduced it into IFCM algorithm to improve these two kinds of defects.However,the existing intuitionistic fuzzy entropy can’t accurately describe the uncertainty and fuzziness of intuitionistic fuzzy sets in some cases,so the effect of its introduction into IFCM algorithm is also very limited.Aiming at the inherent defects of intuitionistic fuzzy entropy,some scholars put forward the intuitionistic fuzzy knowledge measure theory,and gradually formed the intuitionistic fuzzy knowledge measure axiom model through continuous development and improvement.This thesis introduces the latest theory of intuitionistic fuzzy knowledge measure into intuitionistic fuzzy clustering algorithm,and proposes two kinds of intuitionistic fuzzy clustering methods based on knowledge measure theory,so as to improve the two kinds of defects of IFCM algorithm.The main research content of this thesis is as follows:1.Aiming at the two kinds of problems of IFCM algorithm,this thesis proposes an intuitive fuzzy mean clustering method based on knowledge weight.Firstly,the data set is intuitively fuzzy and the corresponding knowledge amount of each feature is calculated to achieve the data feature weighting.Then,the kernel space density of the weighted data is calculated and the clustering center is initialized with the maximum and minimum method,so as to improve the clustering efficiency and calculation accuracy.Finally,based on the latest intuitionistic fuzzy knowledge measure theory,an improved intuitionistic fuzzy knowledge measure formula was proposed,which was introduced into the objective function of clustering,and a new membership formula was derived to improve the accuracy and efficiency of clustering.2.Aiming at the weight calculation process of the missing similarity of the existing intuitionistic fuzzy sets,this thesis constructs a similarity degree from the perspective of the amount of information and the clarity of information in the intuitionistic fuzzy knowledge measure theory,and proposes a weighted similarity degree of intuitionistic fuzzy sets based on the knowledge weight method.Finally,the weighted similarity degree is used to construct a similarity matrix,and an intuitionistic fuzzy clustering method based on knowledge weighted similarity matrix is proposed.This thesis introduces the new theory of knowledge measure into the intuitionistic fuzzy clustering method for the first time and achieves good results.It creates a new example for the potential application of this theory in other related fields.