To Improve the Classification Accuracy by Using the MDESM Based Classification Approach

Abstract

The problem of classification has been discussed in many situations and there are variety of
methods has been proposed by the researchers to improve the performance of classification
but suffers with the problem of poor classification accuracy. To solve these issues, we
propose a multi-dimensional eccentric similarity measure based classification algorithm,
which computes eccentric similarity between data points of each class at each dimension
rather than computing distance with the points near to the center of the cluster. The method
computes eccentric similarity measure at each dimension and finally computes a cumulative
similarity with each cluster before assigning a label to the data point. The eccentric similarity
measure is the distance between the data points which are located at the boundary of any
cluster and we compute both the distance from the all the data points and the distance from
the boundary points. By classifying data points based on eccentric similarity measure, the
quality of classification has been improved and false indexing has been reduced.