An Improved Clustering Analysis Method Based on Fuzzy C-Means Algorithm by Whale Optimization Algorithm

The use of big data has become widespread in many areas of human knowledge, including medicine and engineering. One of the most widely used processes on various types of data, especially big data, is cluster analysis or clustering. One of the most popular clustering methods developed by the distance approach is called Fuzzy C-Means (FCM). Having a simple structure, this method has been favored by developers in many applications; however, it cannot be used in the clustering of big data. Due to the large number of objects, they are not loadable in the main memory of ordinary computer systems at run time; therefore, they are impossible to be processed at once. Moreover, FCM is sensitive to cluster center initialization, so that inappropriate initialization may lead to slow or non-optimal convergence. Optimization methods are usually used to solve the FCM convergence problem and to find more appropriate cluster centers. In this thesis, a new clustering method has been introduced in which a whale optimization algorithm is used to solve the FCM convergence problem. Furthermore, random sampling of data, application of the clustering on samples, and ultimately, extension of the clustering results to all data have been proposed as a solution for the problem of big data clustering. In order to reduce the effect of the selected samples on the performance of clustering in this solution, sampling is repeated several times, and at the end, the clustering results are combined. Results from the application of the proposed clustering method on artificial and actual databases indicate the accuracy of the proposed method compared with that of other similar methods.

Keywords:

Fuzzy c-means clustering , Whale optimization , FCM , Data clustering , Big Data