A new algorithm for data clustering using combination of genetic and Fireflies algorithms
Subject Areas : Multimedia Processing, Communications Systems, Intelligent Systems
Mahsa Afsardeir
1
(Islamic Azad University, Science and Research Branch, Faculty of Engineering, Department of Computer Engineering, Tehran, Iran.)
mansoure Afsardeir
2
(Islamic Azad University, Dezful Branch, Faculty of Engineering, Department of Biomedical Engineering, Dezful, Iran.)
Keywords: firefly algorithm.Firefly Genetic Algorithm. k-means algorithm, Data mining, genetic algorithms,
Abstract :
Introduction: With the progress of technology and increasing the volume of data in databases, the demand for fast and accurate discovery and extraction of databases has increased. Clustering is one of the data mining approaches that is proposed to analyze and interpret data by exploring the structures using similarities or differences. One of the most widely used clustering methods is the k-means. In this algorithm, cluster centers are randomly selected and each object is assigned to a cluster that has maximum similarity to the center of that cluster. Therefore, this algorithm is not suitable for outlier data since this data easily changes centers and may produce undesirable results. Therefore, by using optimization methods to find the best cluster centers, the performance of this algorithm can be significantly improved. The idea of combining firefly and genetics algorithms to optimize clustering accuracy is an innovation that has not been used before.Method: In order to optimize k-means clustering, in this paper, the combined method of genetic algorithm and firefly worm is introduced as the firefly genetic algorithm.Findings: The proposed algorithm is evaluated using three well-known datasets, namely, Breast Cancer, Iris, and Glass. It is clear from the results that the proposed algorithm provides better results in all three datasets. The results confirm that the distance between clusters is much less than the compared approaches.Discussion and Conclusion: The most important issue in clustering is to correctly determine the cluster centers. There are a variety of methods and algorithms that performs clustering with different performance. In this paper, based on firefly metaheuristic algorithms and genetic algorithms a new method has been proposed for data clustering. Our main focus in this study was on two determining factors, namely the distance within the data cluster (distance of each data to the center of the cluster) and the distance that the headers have from each other (maximum distance between the centers of the clusters). In the k-means algorithm, clustering is not accurate since the cluster centers are selected randomly. Employing firefly algorithms and genetics, we try to obtain more accurate centers of the clusters and, as a result, correct clustering.
_||_