A multi-constraint based objective function and lion optimization for the data clustering

Data clustering allows partitioning of the large database into smaller databases for improving the data mining. This paper introduces a multi-constraint based objective function for the selection of the optimal cluster centroids for clustering the data. The proposed objective function utilizes three parameters, such as intra-cluster distance, inter-cluster distance, and density, where various kernel functions, namely Gaussian kernel, tangential kernel, rational quadratic kernel and inverse multiquadratic kernel, are used for the distance measure. Then, Adaptive Dynamic Directive Operative Fractional Lion (ADDOFL) algorithm that makes use of proposed multikernel based objective function finds the optimal cluster center. From the simulation results, it is obvious that the ADDOFL algorithm with the proposed objective function has better performance than the ADDOFL algorithm, using the Clustering Accuracy (CA) and the Jaccard coefficient (JC) metrics. For the lung cancer database, the proposed ADDOFL with the multikernel based fitness achieved CA and JC values of 0.8670, and 0.85, for the cluster size of 2. Similarly, for the Pima Indian diabetes database, the proposed ADDOFL with the multikernel based fitness achieved CA and JC values of 0.84609, and 0.85, respectively.