Multi kernel and dynamic fractional lion optimization algorithm for data clustering

• Published in: Alexandria engineering journal Vol. 57; no. 1; pp. 267 - 276
• Main Authors: Chander, Satish, Vijaya, P., Dhyani, Praveen
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2018; Elsevier
• Subjects: Clustering accuracy; Data clustering; Directive operative searching strategy; Fractional lion optimization; Multi kernel function; Fractional lion optimization; Data clustering; Multi kernel function; Directive operative searching strategy; Clustering accuracy
• ISSN: 1110-0168; 2090-2670
• DOI: 10.1016/j.aej.2016.12.013

Clustering is the technique used to partition the homogenous data, where the data are grouped together. In order to improve the clustering accuracy, the adaptive dynamic directive operative fractional lion algorithm is proposed using multi kernel function. Also, we intend to develop a new mathematical function for fitness evaluation. We utilize multi kernels, such as Gaussian, tangential, rational quadratic and Inverse multiquadratic to design the new fitness function. Consequently, the WLI fuzzy clustering mechanism is employed in this paper to determine the distance measurement based on new fitness function, named Multi kernel WLI (MKWLI). Then, we design a novel algorithm with the aid of dynamic directive operative searching strategy and adaptive fractional lion algorithm, termed Adaptive Dynamic Directive Operative Fractional Lion (ADDOFL) algorithm. Initially in this proposed algorithm, the solutions are generated based on the fractional lion algorithm. It also exploits the new MKWLI fitness function to evaluate the optimal value. Finally, the updation of female lion is performed through dynamic directive operative searching algorithm. Thus, the proposed ADDOFL algorithm is used to find out the optimal cluster center iteratively. The simulation results are validated and performance is analyzed using metrics such as clustering accuracy, Jaccard coefficient and rand coefficient. The outcome of the proposed algorithm attains the clustering accuracy of 89.6% for both Iris and Wine databases which ensures the better clustering performance.