I have run a kmeans procedure and got the cluster centroids. Now, i want to calculate the euclidean distance from each row to its respective cluster centroids. In other words, i need to calculate the within cluster distances. Any ideas?
What i can of think is:
For simplicity, assuming only one cluster, 1. I will merge(combine rows) the exampleset with the cluster centroid exampleset. And do the 'ExampleSet2Similarity' operator. And then have to delete many rows(unncessary calculations) in the resultant exampleset. This complicates when the number of clusters are many and for large data too. 2. Merge(combine columns) of the exampleset with the corresponding cluter centroids (this may take some time). And then have to actually compute the formula for euclidean distance.
But i find both of these methods as tedious and was wondering if there was a simpler solution for this problem.
Hi, yes there is a much simpler solution: The operator ClusterCentroidEvaluator provides the possibility to calculate the average squared within distance automatically, given an exampleSet and a cluster model.
Thanks Sebastian, But i dont want the average. I need for each of the observations in the ExampleSet, to calculate the euclidean distance from its respective centroids. So, a column will be generated in the ExampleSet, giving the euclidean distance.