Абстрактный

A New Algorithm for Finding Automatic Clustering In Unlabeled Datasets

S Chitra, G Komarasamy

Data mining refers to extracting or mining knowledge from large amount of data. In these data mining has different models, clustering is used as the descriptive type model. Clustering is task of grouping a set of physical or abstract objects into classes of similar objects. Clustering is also referred to as unsupervised learning or segmentation.In this clustering techniques k-means clustering algorithm is a vital role to group the objects. In this algorithm user can give the number of clusters in priori as k value. The k value depends on the final clustering objects, to avoid such a problem proposed the new Multi Objective (MO) clustering technique with combined form of Genetic clustering MO Optimization (GenClustMOO) and Archived Multi Objective Simulated Annealing (AMOSA) used for finding an automatic k value as the best center point. The proposed method is global search and local search are combined which improving the performance. The datasets are taken from UCI repository for verify the performance of the algorithm.