MyKmeans
代码说明:
实现聚类K均值算法: K均值算法:给定类的个数K,将n个对象分到K个类中去,使得类内对象之间的相似性最大,而类之间的相似性最小。 缺点:产生类的大小相差不会很大,对于脏数据很敏感。 改进的算法:k—medoids 方法。这儿选取一个对象叫做mediod来代替上面的中心 的作用,这样的一个medoid就标识了这个类。步骤: 1,任意选取K个对象作为medoids(O1,O2,…Oi…Ok)。 以下是循环的: 2,将余下的对象分到各个类中去(根据与medoid最相近的原则); 3,对于每个类(Oi)中,顺序选取一个Or,计算用Or代替Oi后的消耗—E(Or)。选择E最小的那个Or来代替Oi。这样K个medoids就改变了,下面就再转到2。 4,这样循环直到K个medoids固定下来。 这种算法对于脏数据和异常数据不敏感,但计算量显然要比K均值要大,一般只适合小数据量。(achieving K-mean clustering algorithms : K-means algorithm : given the number of Class K, n will be assigned to target K to 000 category, making target category of the similarity between the largest category of the similarity between the smallest. Disadvantages : class size have no great difference for dirty data is very sensitive. Improved algorithms : k-medoids methods. Here a selection of objects called mediod to replace the center of the above, the logo on a medoid this category. Steps : 1, arbitrary selection of objects as K medoids (O1, O2, Ok ... ... Oi). Following is a cycle : 2, the remaining targets assigned to each category (in accordance with the closest medoid principle); 3, for each category (Oi), the order of selection of a Or, calculated Oi Or replace the consumption-E (Or))
下载说明:请别用迅雷下载,失败请重下,重下不扣分!