ward's method is a method of hierarchical clustering of Law distance between the common cluster metric aggregation, which is a basic process (Reference: http://blog.sciencenet.cn/blog-2827057-921772.html )
-
ESS is calculated for each cluster of
-
Calculate the total ESS
-
Enumerate all two cluster is cluster of N [N * (N-1) / 2] th two sets, ESS total calculated value to merge the two cluster
-
Select ESS total value of the minimum growth that two cluster merge
-
The process is repeated until N is reduced to 1
In fact, this method is relatively inefficient, especially ESS count value but also the first cluster of averaging points, then counted again summed square of the distance, but there is a quick calculation method called Lance-Williams Algorithm can greatly simplify the calculation of the ward method. First to a chart (Source: https://www.youtube.com/watch?v=aXsaFNVzzfI