Subtractive clustering for categorical data
Lei Gu · 2016
Previous subtractive clustering methods can be used for numerical data, but it cannot be applied to categorical data because attribute values of categorical data do not have a natural ordering. In this paper, one novel subtractive clustering method which is applied to some categorical data is given. The Euclidean distance is replaced by Hamming distance in this new approach. Some experiments in this paper are run on several UCI datasets, and some experimental results describe that this novel presented method can get the better clustering accuracies compared to traditional k-modes.