ost-processing Two post-processing strategies that decrease the Sse by increasing the number of clusters are Split a cluster The cluster with the largest SSE is usually chosen Introduce a new cluster centroid Often the point that is farthest from its associated cluster center is chosen We can determine this if we keep track of the contribution of each point to the sse 8/20/2016 PATTERN RECOGNITIONPost-processing Two post-processing strategies that decrease the SSE by increasing the number of clusters are ◦ Split a cluster ◦ The cluster with the largest SSE is usually chosen. ◦ Introduce a new cluster centroid ◦ Often the point that is farthest from its associated cluster center is chosen. ◦ We can determine this if we keep track of the contribution of each point to the SSE. 8/20/2016 PATTERN RECOGNITION 31