Optimal data clustering by using artificial immune network with elitist learning

Data clustering is one of the most popular techniques in data mining. The attributes of the objects within the same cluster should be similar to each other, while the attributes of the objects within different clusters should be different. Most of classical clustering algorithms suffer from such disadvantages as local optima, initial centroid selection and the lack of a way to automatically determine the optimal number of clusters, and thus fail to cluster datasets with complex distributions. This paper proposes a novel clustering algorithm based on artificial immune network with elitist learning (Ocopt-aiNet-EL). Some elaborated operators are implemented for automatic data clustering and the Manifold distance is used for the dissimilarity measure of antibodies. A number of benchmark cases including artificial datasets and real-world datasets are arranged to evaluate the performance of the proposed Ocopt-aiNet-EL algorithm and the other clustering algorithms, such as K-means, EAC and Ocopt-aiNet. The experimental results demonstrate that the proposed Ocopt-aiNet-EL is an effective and efficient clustering method.