Attribute weighted fuzzy clustering algorithm based on mutual information

Abstract

It is studied by applying the mutual information which is used to assess the contribution of each attribute that has the different important degrees to the classification in the fuzzy clustering algorithm, then the attribute weighted fuzzy clustering algorithm based on mutual information is proposed. By using the mutual information to quantify the contribution of each attribute to the classification, the attributes are weighted and introduced into the fuzzy C mean algorithm. For incomplete data sets, the missing attribute is also introduced as a target object to be optimized and as a part of the iterative to be optimization. Finally, an example verifies the applicability of the algorithm in dealing with incomplete data sets and incomplete data sets, and analyzes the effect of each attribute value loss on clustering results in incomplete data sets.