Moth flame optimization algorithm based on quadratic interpolation for data clustering

Abstract

K-means clustering is a clustering technique based on partition. It is widely used in practice because of its simplicity and efficiency. However, it has shortcomings of highly relying on initial clustering center and possibly trapping into local optimum. Firstly, a revised moth flame optimization algorithm with quadratic interpolation is proposed to overcome the defects of K-means and to revise the quality of solution and iterative efficiency of the basic algorithm. The initial population with better diversity is generated by using tent chaotic map to ameliorate the exploration ability of the algorithm. Arithmetical crossover operation for flame location produces new flame with better diversity to guide the moth finding the optimal solution so that the iterative efficiency of the algorithm is ameliorated. Selecting the moths of population to perform quadratic interpolation is helpful for the algorithm to converge rapidly near the optimal solution. It can polish up exploitation ability of the algorithm. The high performance of the improved algorithm is then employed for optimize the location of cluster centers and is examined by five UCI datasets. The experiment results indicate that the improved technique is suitable for finishing problem clustering via k-means, and good clustering results are obtained.