Attributes Scaling for K-means Algorithm Controlled by Misclassification of All Clusters

摘要

K-means clustering is one of the well-known distance-based clustering methods which partitions data into distinct groups. To implement an automatic attribute-scaled K-mean algorithm, the concept of classification has been integrated. Data points which belong to the same target class are considered similar in K-means clustering. In this paper, we explore and determine the optimal attribute-scaled vector that minimizes misclassification error of the target class. This paper uses the non-linear unconstrained optimization technique in attribute-scaled space, called the cyclic coordinate method together with the golden section line search to find the optimal vector. Our experiments show that the methods can provide the optimal scaling vectors which effectively reduce the misclassification error of supervised K-means clustering and lead to the effective supervised clustering in some data sets.