An Improved Crow Search Algorithm for Data Clustering

Metaheuristic algorithms are often trapped in local optimum solutions when searching for solutions. This problem often occurs in optimization cases involving high dimensions such as data clustering. Imbalance of the exploration and exploitation process is the cause of this condition because search agents are not able to reach the best solution in the search space. In this study, the problem is overcome by modifying the solution update mechanism so that a search agent not only follows another randomly chosen search agent, but also has the opportunity to follow the best search agent. In addition, the balance of exploration and exploitation is also enhanced by the mechanism of updating the awareness probability of each search agent in accordance with their respective abilities in searching for solutions. The improve mechanism makes the proposed algorithm obtain pretty good solutions with smaller computational time compared to Genetic Algorithm and Particle Swarm Optimization. In large datasets, it is proven that the proposed algorithm is able to provide the best solution among the other algorithms.