Research And Apply Tothe Decision Graph Algorithm

With the era development, especially in last few years into the data explosion era, the importance of data mining increasingly highlights.And some of the classic algorithms have been unable to meet the growing demand for data processing.Cluster, as an important field in data mining, demand for better clustering algorithms is also increasingly urgent.Decision Graph (DG) clustering algorithm (is also called CFSFDP algorithm), which published in June 2014 in the journal Science has been widely discussed. The clustering algorithm has some good properties. Moreover the idea of the algorithm on the selection of clustering center is very concise and clever.Some think highly of the method of selecting the clustering center. Others don't like the selection of the clustering center because it requires human participation. However, the author has a positive attitude to the algorithm. After all, the DG algorithm has obvious advantages, and the disadvantages are not steady.After being extensive optimizationed and improved,the DG algorithm can become a classic algorithm in the field of clustering, too.A new problems of the DG algorithm will be presented in this paper, in which the solutions will be presented, too.The main content of this paper is aimed at the found problem of the algorithm design:conflicts of density. Then to solve the problem, two improvements have been designed and implemented. The algorithm is verified by comparative experiment of the mistakes and the effect of the solution of the improved algorithm.Finally, decision graph algorithmis used on the simple classic K-means algorithm, for proving the advantages of the method choosing the clustering center of the DG algorithm.New algorithm, which was called DG-means, is designed and implemented.New algorithm is verified by comparing with K-means algorithm to be proved that it has all kinds of excellent properties.Indirectly proves that the DG algorithm in selecting the good effect of clustering centers.