A hybrid strategy for krill herd algorithm with harmony search algorithm to improve the data clustering

Krill herd (KH) is a stochastic nature-inspired optimization algorithm, it has been successfully used to solve many involved optimization problems. Occasionally, poor exploration (diversification) capability affects the performance of krill herd algorithm (KHA). In this paper, we proposed a new hybridization strategy, namely, hybrid the krill herd algorithm with the harmony search (HS) algorithm (Harmony-KHA), to improve the data clustering technique. This hybridization strategy seeks to enhance the global (diversification) search capability of the KH algorithm to obtain the best global optima. The proposed algorithm are conducted through the addition of the global search operator from the HS algorithm in order to improve the exploration around the optimal solution in KH and thus kill individuals move towards the global best solution. The proposed algorithm is applied to keep the best krill individuals during the updating positions of the krill individuals. Experiments were conducted using four standard datasets from the UCI Machine Learning Repository, which is commonly used in the domain of data clustering. The results showed that the proposed hybrid the KH algorithm with the HS algorithm (Harmony-KHA) is produced very accurate clusters especially in the large dataset. Furthermore, the Harmony-KHA obtained a high convergence rate and it can overcome the other comparative algorithms. The proposed algorithm is compared with other well-known based on data clustering algorithms including the original KH algorithm.