Vinayaka : A Semi-Supervised Projected Clustering Method Using Differential Evolution

Satish Gajawada

doi:10.5121/ijsea.2012.3406

Abstract

Differential Evolution (DE) is an algorithm for evolutionary optimization. Clustering problems have been solved by using DE based clustering methods but these methods may fail to find clusters hidden in subspaces of high dimensional datasets.Subspace and projected clustering methodshave been proposed in literature to find subspace clusters that are present in subspaces of dataset. In this paper we propose VINAYAKA, a semi-supervised projected clustering method based on DE. In this method DE optimizes a hybrid cluster validation index. Subspa ce Clustering Quality Estimate index (SCQE index) is used for internal cluster validation and Gini indexgain is used for external cluster validationin the proposed hybrid cluster validation index.Proposed method is applied on Wisconsin breast cancer dataset.

Highlights

Differential Evolution (DE) was proposed by Price and Storn [1] which is a evolutionary based optimization technique and based on a differential operator
The clustering solutions obtained for various values of number of clusters parameter are validated with a hybrid cluster validation index and an internal validation index
In this paper we proposed VINAYAKA, a semi-supervised projected clustering method using Differential Evolution optimization technique

Summary

INTRODUCTION

Differential Evolution (DE) was proposed by Price and Storn [1] which is a evolutionary based optimization technique and based on a differential operator. External cluster validation indices use external information that is available about the data [7]. There exists no best cluster validation index which always gives better result compared to other indices. Better results can be obtained by fusion of various cluster validation indices compared to using single cluster validation index for getting optimal clustering solution. Internal validation indices like DaviesBouldin index and Dunn index can be fused to validate clustering solutions for obtaining optimal clustering solution [13]. Using fusion of internal validation indices can give better results but available external information about the dataset is not used in the validation of clustering solution. In this paper we propose a hybrid cluster validation index for high dimensional datasets using SCQE index for internal cluster validation and Gini index gain for external cluster validation.

RELATED WORK

PROPOSED WORK

EXPERIMENTAL RESULTS

CONCLUSIONS