Authors : K.S. Kavitha, K.V. Ramakrishnan and Manoj Kumar Singh
Abstract: In this study, a new approach has applied to define the clustering using factorizing the original data set matrix into two lower dimension matrices namely, two dimensional features data set and a transformation matrix with the help of non negative matrix factorization. This two dimensional feature data set is having the more separation in available different categories and also provide approximated visual information about possible clusters available in data set along with correlation available among them. Two dimension feature sets are a used to obtain the final clusters using optimizing the minimum quantizing error with help of Genetic algorithm. Comparisons are made with other well established algorithms like particle swarm optimization. Benefits of features matrix is also shown in compare to raw data set in terms of obtained cluster performance. K-means algorithm is also applied independently before and after matrix factorization and comparisons are made with other obtained results. Cluster performance indexes are defined in terms of F-measure and purity.
K.S. Kavitha, K.V. Ramakrishnan and Manoj Kumar Singh, 2014. Data Clustering using GA with Non-Negative Matrix Factorization. Asian Journal of Information Technology, 13: 222-234.