Designing a New Hybrid K-Means Optimization Algorithm

Abstract

Abstract- Clustering is an unsupervised learning process with the main objective of organizing data into certain clusters and groups such that the data objects in the same cluster have higher similarity but dissimilar to the objects in other clusters. Many algorithms have been presented till now. These algorithms are k-means algorithm, hybrid k-means algorithms, and variants of k-means. The hybrid k-means optimization algorithms are those that combines k-means algorithm with various optimization algorithms like KBAT, CGA, KFA etc. The k-means algorithms have many problems like k-means algorithm converges to local minima rather than giving a global optimum results, hybrid k-mean algorithms have accuracy and efficiency problems. A new hybrid k-means optimization clustering algorithm is presented in the proposed study to solve the problems of existing k-means algorithm and hybrid k-mean algorithms. The output parameters like intra cluster distance, purity, recall etc are use to compare the performance of existing hybrid k-means algorithm with a new hybrid k-means optimization clustering algorithm. The results of these output parameters show that the proposed ABCGA technique is more efficient and it requires less computational time than existing hybrid k-means algorithm. Also the recall and purity of proposed hybrid algorithm is better than existing hybrid k-means algorithms and k-means algorithm but still there is scope for improvement in the proposed technique to extend this approach to handle variety of situations and information.