| With the development of technology and society, more and more things havecome out, and consequently, the number of varieties in our world is getting larger. Inthis situation, if we can cluster things into many fewer varieties, we can learn thingsfaster and easier. Since the first time people began to apply cluster methods intopractice, many different means have been invented. And, many flaws in the past havebeen eliminated due to the endeavor of hard-working researchers.But now, I see another problem coming. With all these methods, is that possiblethat we can put some together because of their some intrinsic similarities? On onehand, a beginner might feel confused when confronted too many means. And on theother hand, even if an expert, who has learned many even a large number of clustermethods, still feel it time-consuming to choose the right one. Can we set somestandards to help choose the right method quickly? In this way, we can save time,money and energy to prompt the further development of our world.Problems mentioned above are concerned in this paper. First, I learnedextensively many different cluster methods, in order to understand the reason,thinking, steps and applications of them. This is the prerequisite of my followingwork.Second, I chose several characteristic methods to represent the many similarmethods behind. Limited by time and energy, I focused on two usual kinds: classiccluster methods and fuzzy cluster methods. This paper also discussed some specificvarieties that belong to them.In the end, according to the preparation before, I concluded four essential factorsthat matter most: the number of the samples that need to be classified, the number ofclusters, the extent of dependence on computer and the particularity of the structure ofthe samples. Then，to testify the effectiveness of the conclusion of this paper, I use anpractical example. The result is quite satisfactory, it successfully prove that at least theconclusion is qualified to some extent. |
PDF Full Text Request