Add files via upload

Author: arnab39

Kmeans_Clustering/Clustering_assignment_2_my_code.py

@@ -0,0 +1,147 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Tue Aug 29 15:31:25 2017
+
+@author: Arnab
+"""
+
+## Initialisation
+
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import scipy.io as sio 
+import copy
+from pprint import pprint
+
+
+
+##For plotting the clusters 
+
+def plot_clusters(df,means,colmap):
+ plt.scatter(df['x'], df['y'], color=df['color'], alpha=0.5, edgecolor='k')
+ for i in means.keys():
+ plt.scatter(*means[i],s=200, color=colmap[i],marker='o')
+ plt.xlim(-1, 5)
+ plt.ylim(-2, 3)
+ plt.show()
+
+
+##Assign mean randomly 
+
+def initial_mean_assignment(df,k):
+ np.random.seed(200)
+ initmeans = {
+ i+1: [np.random.uniform(-2,4), np.random.uniform(-1, 2)]
+ for i in range(k)
+ }
+ return initmeans
+
+## Distance Function
+
+def distance(x1,y1,x2,y2,n):
+ if n==0:
+ return np.sqrt( (x1-x2)**2 + (y1-y2)**2 )
+ elif n==1:
+ return 1-(x1*x2+y1*y2)/(np.sqrt(x1**2+y1**2)*np.sqrt(x2**2+y2**2))
+
+
+## Assignment Stage
+
+def One_iteration_clustering(df, centroids, colmap,dist_m):
+ for i in centroids.keys():
+ df['distance_from_{}'.format(i)] = distance(df['x'],df['y'],centroids[i][0],centroids[i][1],dist_m)
+ mean_to_distance_cols = ['distance_from_{}'.format(i) for i in centroids.keys()]
+ df['closest'] = df.loc[:, mean_to_distance_cols].idxmin(axis=1)
+ df['closest'] = df['closest'].map(lambda x: int(x.lstrip('distance_from_')))
+ df['color'] = df['closest'].map(lambda x: colmap[x])
+ 
+ return df
+ 
+## Update the mean values 
+
+def update_means(df,means):
+ for i in means.keys():
+ means[i][0] = np.mean(df[df['closest'] == i]['x'])
+ means[i][1] = np.mean(df[df['closest'] == i]['y'])
+ return means 
+ 
+ 
+## For clusting 
+
+def kmeans_clustering(df,k,colmap,dist_m):
+ 
+ means=initial_mean_assignment(df,k)
+ i=0;
+ while 1:
+ old_means=copy.deepcopy(means)
+ df = One_iteration_clustering(df, means, colmap,dist_m)
+ means = update_means(df,means)
+ i=i+1;
+ print "Iteration"+str(i)
+ if old_means==means:
+ break
+ return means 
+ 
+#To compare the two distance metrics 
+
+def cluster_comparision(df,colormap):
+ 
+ eeuc=[]
+ ecos=[]
+ k=[2,3,4,5,6]
+ er_euc = pd.DataFrame({'2': [0],'3': [0],'4': [0],'5': [0],'6': [0]})
+ er_cos = pd.DataFrame({'2': [0],'3': [0],'4': [0],'5': [0],'6': [0]})
+ for i in range(2,7):
+ mean_euc=kmeans_clustering(df,i,colmap,0)
+ for j in range(len(df.index)):
+er_euc['{}'.format(i)] += df['distance_from_{}'.format(df['closest'][j])][j]
+ mean_cos=kmeans_clustering(df,i,colmap,1)
+ for j in range(len(df.index)):
+er_cos['{}'.format(i)] += df['distance_from_{}'.format(df['closest'][j])][j]
+ eeuc.append(er_euc['{}'.format(i)])
+ print er_euc['{}'.format(i)]
+ ecos.append(er_cos['{}'.format(i)])
+ print er_cos['{}'.format(i)]
+ plt.plot(k,eeuc)
+ plt.plot(k,ecos)
+ plt.xlabel('Number of Clusters')
+ plt.ylabel('Error')
+ plt.title('Error variation with number of clusters')
+ return df
+ 
+## To illustrate kmeans clustering 
+ 
+def kmean_illus_(df,k,colmap,dist_m):
+ means=kmeans_clustering(df,k,colmap,dist_m)
+ plot_clusters(df,means,colmap)
+ print "Successfully clustered with k="+str(k) 
+ 
+ 
+if __name__ == '__main__': 
+ a = sio.loadmat('data.mat')
+ data=a['h']
+ #pprint(data)
+ df = pd.DataFrame(data,columns=list('xy'))
+ colmap = {1: 'r', 2: 'g', 3: 'b', 4: 'c', 5: 'm', 6: 'y', 7: 'k'}
+ 
+ '''
+ Code segment to see the illustration of the kmeans clustering 
+ dist_m values selects which distance metric we want to use 
+ dist_m = 0 for Euclidean distance 
+ dist_m = 1 for Cosine distane 
+ '''
+ dist_m = 1
+ k=3
+ #kmean_illus(df,k,colmap,dist_m)
+ 
+ '''
+ Code segment to see the plot of the error with number of clusters
+ Knee point observed is k=3 
+ '''
+ 
+ df=cluster_comparision(df,colmap)
+ 
+ 
+
+