Java: Data Science Made Easy

K-means

As we mentioned previously, Agglomerative clustering methods work quite well with small datasets, but they have some problems with bigger ones. K-means is another popular clusterization technique, which does not suffer from this problem.

K-means is a clustering method, which belongs to the partitioning family of clustering algorithm: given the number of clusters K, K-Means splits the data into K disjoint groups. Grouping items into clusters is done using centroids. A centroid represents the "center" of a cluster, and for each item, we assign it to the group of its closest centroid. The quality of clustering is measured by distortion - the sum of distances between each item and its centroid.  

As with agglomerative clustering, there ...