kmeans.py

#Written by Metehan Ozten 
# use cmdline arg '-i' to specify input file
# use cmdline arg '-v' to turn on verbose printing

import matplotlib.pyplot as plt
import numpy
import scipy.cluster.hierarchy as hcluster
import sys
import scipy.cluster.vq as kclust
def print_centroids(clusters):
	print('Data Columns: public_favorites_count|followings_count|followers_count|likes_count|track_count|playlist_count|comments_count')
	for x in range(0, clusters.shape[0]):
		print('Centroid %d: %s' % (x, str(clusters[x])))


def print_data_with_labels(data, analysis_data, labels):
	for i in range(0, data.shape[0]):
		print('user_id: %d id: %d %s label: %d' % (analysis_data[i][0], analysis_data[i][1], str(data[i]), labels[i]))


def main():
	use_cols = (2,3,4,5,6,7,8)
	if '-i' not in sys.argv:
		analysis_data = numpy.loadtxt('features.txt', skiprows = 1, delimiter = ',')
		data = numpy.loadtxt('features.txt', skiprows = 1, usecols = use_cols, delimiter = ',')
	else :
		analysis_data = numpy.loadtxt(sys.argv[sys.argv.index('-i')+1], skiprows = 1, delimiter = ',')
		data = numpy.loadtxt(sys.argv[sys.argv.index('-i')+1], skiprows = 1, usecols = use_cols, delimiter = ',')
	num_clusters = 4 #default
	if '-n' in sys.argv:
		num_clusters = int(sys.argv[sys.argv.index('-n')+1])
	clusters, labels = kclust.kmeans2(data, num_clusters)
	title = "number of clusters: %d" % (len(clusters))
	print_centroids(clusters)
	if '-v' in sys.argv:
		print_data_with_labels(data, analysis_data, labels)
	
if __name__ == '__main__':
	main()