Optimized cluster validation technique for unsupervised clustering techniques
R. Krishnamoorthy, Sushant Kumar · 2014
In this paper, a new cluster validation technique called Optimized Cluster Validation (OCV) is presented. The proposed technique is aimed to measure the purity and impurity over the resulting cluster of the unsupervised clustering techniques. The proposed OCV technique consists of two measures which are Purity Measure (PM) and Impurity Measure (IM). The first measure (PM), is aimed to measure the intra cluster similarity or intra cluster purity, and it evaluates the overall resulting cluster quality or accuracy or purity. The second measure (IM), is evaluate the intra cluster dissimilarity or intra cluster impurity over the resulting cluster of the unsupervised clustering technique. The experimental results show that the OCV technique is simple and effective to evaluate the intra cluster similarity and dissimilarity around the resulting cluster of the unsupervised clustering techniques.