python 2.7 - Percentage of proximity to cluster in kmeans -
i trying clustering in python.
this script follows:
import scipy import numpy np scipy.cluster.vq import kmeans2 collections import counter def dist(a,b): return np.linalg.norm(a-b) x = scipy.randn(100, 2) k=int(raw_input("give value of k: ")) pts = 100*np.random.random((12,2)) centroids, assigned_clusters = kmeans2(x, k) pt in pts: #print tuple(pt) distperxy=dict() i=1 cent in centroids: distperxy[i]=(float(dist(pt,list(cent)))) i+=1 mnmm=min(distperxy.iterkeys(), key=(lambda key: distperxy[key])) tot = sum(distperxy.values()) perc=distperxy[mnmm]/tot print mnmm,perc
basically trying determine nearest cluster each point , percentage of proximity cluster. example if point near centroid1 in k-means k=3. percentage calculation described in code above,i.e, dividing distance closest cluster sum of distances clusters?
Comments
Post a Comment