A Noise Insensitive Cluster Validity Measure for Pattern Classification

When trying to discover knowledge on a collection of data, one of the first arising tasks is to identify groups of similar objects, that is, to carry out cluster analysis for obtaining data partitions. Thus, a decision must be taken for choosing the clustering result that produces the best data partition for a given data collection. In order to support such a decision, indexes for measuring the quality of a data partitioning must be constructed. So far, several cluster validity indexes have been formulated in the literatures. Each of those indexes has strengths and drawbacks when compared with the others. In the present study, an alternative cluster validity index is formulated. The proposed validity index not only takes the contribution of each pattern into consideration, but also relies on information of intra-cluster and inter-cluster distance. The main advantage of the presented index is that is insensitive to noise by introducing the Gaussian kernel into the proposed validity index. An experimental design was devised in order to determine the comparative performance of the proposed cluster validity index against DB index previously formulated in the literature. Experimental results show that the proposed index is insensitive to noise and adaptive to produce good clustering solution.