The Reasearch Of Gravitational Fuzzy Clustering Algorithm

Today is the era of big data, how to find new knowledge from these big data is an urgent task, and therefore data mining technology gradually rise and become a mainstream technology. Cluster analysis is an important area of data mining, Clustering algorithm is unsupervised learning, Clustering algorithm without any a priori knowledge. Its purpose is to divide the data set into a number of target classes, making the similarity of data objects within the same class as high as possible, and the difference of different classes as high as possible. Cluster analysis already quite mature, clustering has been widely used in data mining, image analysis and pattern recognition and many other areas.This dissertation introduces the theoretical knowledge of clustering and summarizes the traditional clustering algorithm. This dissertation mainly studies fuzzy C-means algorithm(FCM), FCM is the most commonly used clustering algorithm, It has a sound theoretical and mathematical foundations. Although the FCM algorithm has been quite outstanding, it also has some disadvantages, so in order to modify these shortcomings, this dissertation proposes a gravitational fuzzy clustering algorithm.FCM algorithm is sensitive to initialization, the initial cluster centers are chosen randomly, this reduces the clustering speed. This dissertation proposes a method for selecting initial centers, it is based on gravity and maximum-minimum distance,it effectively reduces the number of iterations and speeds up the clustering speed. FCM algorithm requires the user to determine the optimal clustering number, this requires the user to have enough experience, this dissertation proposes a cluster validity index to automatically determine the optimal clustering number.this validity index simulates Xie-Beni index, not based on distance but on gravity, And it achieves good results.The objective function of FCM algorithm is based on the distance,the GFCM is based on first cosmic velocity. The first cosmic velocity is the speed with which an object needs to be traveling to escape from a center. The greater the speed it is more difficult to escape,this is very similar to the cluster membership, this reflects the natural essence of the clustering process.Finally, this dissertation through the experiment verifies the GFCM algorithm,GFCM algorithm can effectively reduce the number of iterations and the ability to determine the optimal clustering number, it improves the quality of clustering, it is feasible and effective.