Clustering

Another technique used in data mining is clustering. SciPy has two modules to deal with any problem in this field, each of them addressing a different clustering tool—scipy.cluster.vq for k-means and scipy.cluster.hierarchy for hierarchical clustering.

Vector quantization and k-means

We have two routines to divide data into clusters using the k-means technique—kmeans and kmeans2. They correspond to two different implementations. The former has a very simple syntax:

kmeans(obs, k_or_guess, iter=20, thresh=1e-05)

The obs parameter is an ndarray with the data we wish to cluster. If the dimensions of the array are m x n, the algorithm interprets this data as m points in the n-dimensional Euclidean space. If we know the number of clusters in which this data should be divided, we enter so with the k_or_guess option. The output is a tuple with two elements. The first is an ndarray of dimension k x n, representing a collection of points—as many as clusters were indicated. Each of these locations...