An FPA and GA-based hybrid evolutionary algorithm for analyzing clusters

摘要

Clustering is a technique employed for data mining and analysis. k-means is one of the algorithms utilized for clustering. However, the answer derived using this algorithm is dependent on the initial solution and hence easily retrieves the optimal local answers. To overcome the disadvantages of this algorithm, in this paper a combination of pollination of flowers algorithm and genetic algorithm, named FPAGA, is presented. Combination algorithms are used to diversify the search space of the solution and to improve its capability. To elaborate, crossover and discarding of pollens operator are utilized to increase the population diversity, while elitism operator is employed to improve the local search capabilities. Five datasets are selected to evaluate the performance of the proposed algorithm. The evaluation results demonstrate not only greater accuracy but also better stability compared to the FPA, GA, FA, DE, and k-means algorithms. Moreover, faster convergence is evident, according to the obtained statistical results.