A novel genetic algorithm-based clustering technique and its suitability for knowledge discovery from a brain data set

Abstract

In this paper we demonstrate that some recent clustering techniques do not produce sensible clusters and fail to discover knowledge from underlying data sets. Sometimes, they obtain a huge number of clusters from few records and sometimes they obtain only two clusters from many records, where one cluster contains one record and the other cluster contains all remaining records. Interestingly, these clustering solutions often achieve high fitness values based on existing evaluation criteria. We in this paper propose a Genetic Algorithm-based Clustering technique called CSClust that produces sensible clusters with high fitness values, which are also useful for knowledge discovery. CSClust learns necessary properties of a sensible clustering solution for a data set from a high-quality initial population, without requiring any user input. It then disqualifies the chromosomes that do not satisfy the properties and replaces them by high-quality chromosomes through its cloning operation in each generation. As a result, it finally produces sensible clusters of high quality that are useful in knowledge discovery from a data set. We apply CSClust on a brain data set and demonstrate its ability in knowledge discovery compared to some existing techniques. We also compare it with five (5) existing techniques on ten (10) publicly available data sets in terms of two well-known evaluation criteria: Silhouette Coefficient and DB Index. Our experimental results demonstrate statistically significant superiority of CSClust over existing techniques.