def main(graphFileName, featuresFileName, featureListFile, trainingFileName, k):
# Read ego network
global graph
graph = igraph.Graph.Read_Ncol(graphFileName)
graph.vs["membership"] = 0
# Read feature names
with open(featureListFile, "r") as featureList:
for line in featureList:
featureNames.append(line.strip())
graph.vs[line.strip()] = None
# Read feature values of nodes
with open(featuresFileName, "r") as featuresFile:
for line in featuresFile:
features = line.split(" ")
for feature in features[1:-1]:
try:
node = graph.vs[int(features[0])]
node[";".join(feature.split(";")[0:-1])] = feature.split(";")[-1]
except:
pass
# Run K-Means on the graph
clusters = kmeans(graph, k)
newClusters = clusters
# Store in each node the cluster it belongs to.
i = 0
for membership in clusters.membership:
graph.vs[i]['membership'] = membership
i = i + 1
# Plot final clusters
print newClusters
igraph.plot(newClusters)
# Plot actual clusters from training data
actualClusters = getTrainingClusters(graphFileName, trainingFileName)
print actualClusters
igraph.plot(actualClusters)
i = 0
count = 0
for node1 in graph.vs:
for node2 in graph.vs:
if(actualClusters.membership[node1.index] == actualClusters.membership[node2.index] and newClusters.membership[node1.index] == newClusters.membership[node2.index]):
count = count + 1
elif(actualClusters.membership[node1.index] != actualClusters.membership[node2.index] and newClusters.membership[node1.index] != newClusters.membership[node2.index]):
count = count + 1
i = i + 1
error = (float(count) / float(i))
print error
def compare(graphFileName, trainingFileName):
graph = igraph.Graph.Read_Ncol(graphFileName)
clustersList = []
actualClusters = getTrainingClusters(graphFileName, trainingFileName)
clustersList.append(graph.as_undirected().community_multilevel())
clustersList.append(graph.community_edge_betweenness())
# clustersList.append(graph.community_optimal_modularity())
clustersList.append(graph.community_label_propagation())
# clustersList.append(graph.as_undirected().community_fastgreedy())
clustersList.append(graph.community_infomap())
clustersList.append(graph.as_undirected().community_leading_eigenvector())
return (actualClusters, clustersList)
def main(graphFileName, featuresFileName, featureListFile, trainingFileName):
# Read ego network
graph = igraph.Graph.Read_Ncol(graphFileName)
# Read feature names
featureNames = []
with open(featureListFile, "r") as featureList:
for line in featureList:
featureNames.append(line.strip())
graph.vs[line.strip()] = None
# Read feature values of nodes
with open(featuresFileName, "r") as featuresFile:
for line in featuresFile:
features = line.split(" ")
for feature in features[1:-1]:
try:
node = graph.vs[int(features[0])]
node[";".join(feature.split(";")[0:-1])] = feature.split(";")[-1]
except:
pass
# Clustering
clusters = graph.community_infomap()
# Store in each node the cluster it belongs to.
i = 0
for membership in clusters.membership:
graph.vs[i]['membership'] = membership
i = i + 1
# Iteratively refine clusters.
igraph.plot(clusters)
parameters = calculateParameters(graph, clusters, featureNames)
newClusters = refineClusters(graph, parameters)
while clusterDifference(clusters, newClusters) == True:
parameters = calculateParameters(graph, newClusters, featureNames)
clusters = newClusters
newClusters = refineClusters(graph, parameters)
# Plot final clusters
print newClusters
igraph.plot(newClusters)
# Plot actual clusters from training data
actualClusters = getTrainingClusters(graphFileName, trainingFileName)
print actualClusters
igraph.plot(actualClusters)
i = 0
count = 0
for node1 in graph.vs:
for node2 in graph.vs:
if(actualClusters.membership[node1.index] == actualClusters.membership[node2.index] and newClusters.membership[node1.index] == newClusters.membership[node2.index]):
count = count + 1
elif(actualClusters.membership[node1.index] != actualClusters.membership[node2.index] and newClusters.membership[node1.index] != newClusters.membership[node2.index]):
count = count + 1
i = i + 1
error = (float(count) / float(i))
print error
