def get_shortest_path(file_path, output_path):
Graph, H = load_graph(file_path)
path_distr = dict()
MxScc = snap.GetMxScc(Graph)
tot = MxScc.GetNodes()
cnt = 0
for NI in MxScc.Nodes():
NIdToDistH = snap.TIntH()
shortestPath = snap.GetShortPath(MxScc, NI.GetId(), NIdToDistH, True)
for ID in NIdToDistH:
dist = NIdToDistH[ID]
if dist in path_distr:
path_distr[dist] += 1
else:
path_distr[dist] = 1
cnt += 1
print '%d/%d' % (cnt, tot)
dataset = list()
for dist in path_distr:
distr = dict()
distr['dist'] = dist
distr['freq'] = path_distr[dist]
dataset.append(distr)
dataset = pd.DataFrame(dataset)
dataset = dataset[['dist', 'freq']]
dataset.sort('dist', ascending=1, inplace=True)
dataset.to_csv(output_path, index=False, encoding='utf-8')
def removeLink(G):
plot = plotting(G, snap.gvlNeato, True)
# 1st layer
plot.hirachical()
print 'DEBUG1', len(plot.community), plot.community
nodeList = []
for node in plot.community:
for NI in G.Nodes():
if snap.GetShortPath(G, NI.GetId(), node) == 1:
if NI.GetId() not in nodeList:
nodeList.append(NI.GetId())
print 'DEBUG2', node, nodeList
nodeList.extend(plot.community)
rmList = snap.TIntV()
for NI in G.Nodes():
if NI.GetId() not in nodeList:
rmList.Add(NI.GetId())
print 'DEBUG3', rmList.Len()
snap.DelNodes(G, rmList)
print '\nRemoving Nodes which is not the adjacent ones of CLAIMID'
print 'Graph has %d Nodes and %d Edges\n' % (G.GetNodes(), G.GetEdges())
return nodeList, rmList
def output_closeness_centrality(graph, filename):
num_nodes = graph.GetNodes()
t0 = time.time()
closeness_dict = {}
for start in graph.Nodes():
# get the sum of shortest path distances from start to all nodes by using snap algorithms
sht_distance_htable = snap.TIntH()
snap.GetShortPath(graph, start.GetId(), sht_distance_htable)
sum_of_sht_paths = 0
for item in sht_distance_htable:
sum_of_sht_paths += sht_distance_htable[item]
closeness_centrality_i = (num_nodes-1)/sum_of_sht_paths
closeness_dict[start.GetId()] = closeness_centrality_i
# sort the closeness centrality values in descending order
closeness_dict = {k:v for k,v in sorted(closeness_dict.items(), key=lambda item: item[1], reverse=True)}
with open(filename, "w") as f:
for i in closeness_dict:
f.write("{} {:.6f}\n".format(i, closeness_dict[i]))
print("Time taken for calculation of closeness centrality = {:.6f}".format(time.time()-t0))
def get_shortest_path(node_id):
NIdToDistH = snap.TIntH()
path_len = snap.GetShortPath(snap_graph, int(node_id), NIdToDistH)
paths = np.zeros((max(node_ids) + 1)) #previously was n_nodes
for dest_node in NIdToDistH:
paths[dest_node] = NIdToDistH[dest_node]
return paths
def analyzeMisc(FNGraph):
# LCC, average distances, clustering
t1 = time.time()
print "Started calculating miscellaneous network statistics:"
print '\tPercentage of nodes in LCC in Football network: %.3f' % (snap.GetMxWccSz(FNGraph) * 100.0)
GraphClustCoeff = snap.GetClustCf (FNGraph, -1)
print "\tClustering coefficient: %.3f" % GraphClustCoeff
diam = snap.GetBfsFullDiam(FNGraph, 1432, False)
print "\tNetwork diameter: %.3f\n" % diam
print "\tCalculating average distance..."
avgDist = 0
iter1 = 0
allNodes1 = FNGraph.GetNodes()
for NI in FNGraph.Nodes():
if(iter1 % 100 == 0):
print "\t\tCalculated for %d nodes" % iter1
NIdToDistH = snap.TIntH()
snap.GetShortPath(FNGraph, NI.GetId(), NIdToDistH)
singleDistSum = 0
for item in NIdToDistH:
singleDistSum += NIdToDistH[item]
avgDist += (1.0/allNodes1) * float(singleDistSum)/(allNodes1-1)
iter1 += 1
print "\tNetwork average distance: %.3f" % avgDist
print "\nFinished calculating in %f seconds\n" % (time.time() - t1)
def make_path_graph(Graphs):
for tup in Graphs:
G, label, color = tup
results = dict()
bigtotal = 0.
bigcount = 0.
for node in G.Nodes():
pathtotal = 0.
count = 0.
for node2 in G.Nodes():
pathtotal += snap.GetShortPath(G, node, node2)
count += 1
if pathtotal / count in results:
results[pathtotal / count] += 1
else:
results[pathtotal / count] = 1
bigtotal += pathtotal
bigcount += pathcount
print(label, bigtotal / bigcount)
x = []
y = []
for key in results:
x.append(key)
y.append(results[key])
inds = np.argsort(x)
x2 = []
y2 = []
for ind in inds:
x2.append(x[ind])
y2.append(y[ind])
plt.loglog(x2, y2, color=color, label=label)
plt.show()
def getDistances(graph):
distances = {}
nameToNId = {}
for n in graph.Nodes():
id = n.GetId()
nameToNId[graph.GetStrAttrDatN(id, 'name').decode('utf-8')] = id
infile = codecs.open('csv/dblpusers.csv', 'r', 'utf-8')
lines = infile.read().splitlines()
dblpUsers = []
for line in lines:
tokens = line.split('||')
if tokens[2] != '' and tokens[1] in nameToNId:
id = int(tokens[0])
distances[id] = {}
dblpUsers.append({'name': tokens[1], 'id': id})
for i in range(len(dblpUsers)):
n1 = dblpUsers[i]['name']
i1 = dblpUsers[i]['id']
for j in range(i + 1, len(dblpUsers)):
n2 = dblpUsers[j]['name']
i2 = dblpUsers[j]['id']
# shortest path behavior is weird if no path exists
dist = snap.GetShortPath(graph, nameToNId[n1], nameToNId[n2])
distances[i1][i2] = dist
distances[i2][i1] = dist
outfile = open('csv/distances.csv', 'w')
for i1 in distances:
for i2 in distances[i1]:
outfile.write(
str(i1) + '||' + str(i2) + '||' + str(distances[i1][i2]) +
'\n')
outfile.close()
infile.close()
return distances
def compute_closeness_centrality(G, GName, Nodes, nodes, edges):
counter = 0
closeness_centralities = []
start_time = time.time()
for NI in G.Nodes():
NIdToDistH = snap.TIntH()
sum_of_shortest_paths = 0
shortestPath = snap.GetShortPath(G, NI.GetId(), NIdToDistH)
for paths in NIdToDistH:
sum_of_shortest_paths = sum_of_shortest_paths + NIdToDistH[paths]
sum_of_shortest_paths = sum_of_shortest_paths + nodes * (
nodes - len(NIdToDistH)) #incorporating unreachable nodes
current_centrality = float(nodes) / sum_of_shortest_paths
closeness_centralities.append([current_centrality, NI.GetId()])
time_taken = time.time() - start_time
print "Execution for Closeness Centrality completed in ", time_taken // 60, " mins and ", (
time_taken // 1) % 60, "seconds"
closeness_centralities.sort()
store_in_file(closeness_centralities, "closeness_centrality", GName)
closeness_centralities.sort(reverse=True)
return closeness_centralities
def get_graph_distance(G, n1, n2, directed=False):
deleted = False
if G.IsEdge(n1, n2):
G.DelEdge(n1, n2)
deleted = True
result = -1 * snap.GetShortPath(G, n1, n2, directed)
if deleted: G.AddEdge(n1, n2)
return result
def avgShortestPath(G): avgPathDir = 0 avgPathUndir = 0 numDirPath = 0 numUndirPath = 0 for src in G.Nodes(): NIdToDistH = snap.TIntH() shortestPathUndir = snap.GetShortPath(G, src.GetId(), NIdToDistH, False) numUndirPath += len(NIdToDistH) for item in NIdToDistH: avgPathUndir += 1.0*NIdToDistH[item]#/len(NIdToDistH) shortestPathDir = snap.GetShortPath(G, src.GetId(), NIdToDistH, True) numDirPath += len(NIdToDistH) for item in NIdToDistH: avgPathDir += 1.0*NIdToDistH[item]#/len(NIdToDistH) print "Avg. Shortest Path (directed): %f"%(1.0*avgPathDir/numDirPath) print "Avg. Shortest Path (undirected): %f"%(1.0*avgPathUndir/numUndirPath)
def getAvgEfficiency(Graph):
nodes = [node.GetId() for node in Graph.Nodes()]
efficiency = 0
n = len(nodes)
for i in range(0, n):
for j in range(i + 1, n):
if i != j:
efficiency += 1 / float(
snap.GetShortPath(Graph, nodes[i], nodes[j]))
return 1 / float(n * n - 1) * efficiency
def getNodeEfficiency(Graph):
nodes = [node.GetId() for node in Graph.Nodes()]
efficiency = []
n = len(nodes)
for i in range(0, n):
for j in range(i + 1, n):
if i != j:
efficiency.append(
1 / float(snap.GetShortPath(Graph, nodes[i], nodes[j])))
return efficiency
def closeness_centrality_node(graph, node, sample=None):
if not sample:
list_nodes = {x for x in range(graph.GetNodes())}
else:
list_nodes = random.sample(xrange(graph.GetNodes()), len(sample))
N = len(list_nodes)
list_of_sp = (snap.GetShortPath(graph, node, x)
for x in list_nodes) # Better to use list comprehension than
# map for clarity and speed reasons
return sum((1.0 / x for x in list_of_sp if x > 0)) / N
def q2_4_aux(name):
G = load_graph(name)
counter = 0.0
for i in range (1000):
path = snap.GetShortPath(G, G.GetRndNId(), G.GetRndNId())
if path != -1:
counter += 1
return counter / 1000
def calc_path_prob(graph):
trials = 1000
reachable_count = 0
for _ in range(trials):
node1 = graph.GetRndNId()
node2 = graph.GetRndNId()
shortestPath = snap.GetShortPath(graph, node1, node2, True)
if shortestPath > 0:
reachable_count += 1
return float(reachable_count) / float(trials)
def ProbabilityOfPath(Graph, nSamples=1000):
'''
Given a graph, returns the sampled probability of two nodes being
connected. Takes nSamples of (u,v) pairs to check for path
connectedness.
'''
paths = 0.0
for _ in xrange(nSamples):
u, v = Graph.GetRndNId(Rnd), Graph.GetRndNId(Rnd)
if snap.GetShortPath(Graph, u, v, True) != NO_PATH: paths += 1
return 100 * paths / nSamples
def getHarmonicClosenessCentr(self, G, nodeId):
n = G.GetNodes()
nodeDistances = snap.TIntH()
snap.GetShortPath(G, nodeId, nodeDistances)
centrValue = 0.0
for nodeKey in nodeDistances:
if (nodeKey != nodeId):
centrValue += (1 / float(nodeDistances[nodeKey]))
centrValue /= (n - 1)
# print nodeId, centrValue
return centrValue
def avg_path_length(self):
""" Brute force average path length calculation """
# TODO: Maybe add dynamic programming to speed up the operation
total_path_length = 0
num_path = 0
for i in range(1, self._num_nodes):
for j in range(i+1, self._num_nodes + 1):
length = snap.GetShortPath(self._graph, i, j)
if length > 0:
num_path += 1
total_path_length += length
else:
pass
return 1.0 * total_path_length / num_path
def TestFracPath(Graph, num_test=1000):
Rnd = snap.TRnd(42)
Rnd.Randomize()
count = 0
i = 0
while i < num_test:
NodeId1 = Graph.GetRndNId(Rnd)
NodeId2 = Graph.GetRndNId(Rnd)
if (NodeId1 != NodeId2):
Length = snap.GetShortPath(Graph, NodeId1, NodeId2, True)
if (Length > 0):
count += 1
i += 1
print(" fraction of reachable pairs", count / num_test)
def GetTrustList(fSourceNode, graph, visitPath, totalNodeList, nodeMapNumDict):
trustList = [0] * len(visitPath)
# tSourceIndex=totalNodeList.index(fSourceNode)
tSourceIndex = nodeMapNumDict[fSourceNode]
# tVisitPathIndex=[totalNodeList.index(node) for node in visitPath]
tVisitPathIndex = [nodeMapNumDict[node] for node in visitPath]
for i in range(len(tVisitPathIndex)):
tempDstNodeIndex = tVisitPathIndex[i]
dist = snap.GetShortPath(graph, tSourceIndex, tempDstNodeIndex)
trustList[i] = 1.0 / dist
return trustList
def harmonic_closeness_centrality():
sizeGraph = UGraph.GetNodes()
NIdToDistH = snap.TIntH()
for node in UGraph.Nodes():
hashTableCount = 0
shortestPath = snap.GetShortPath(UGraph, node.GetId(), NIdToDistH)
for x in NIdToDistH:
if (NIdToDistH[x] != 0):
hashTableCount += float(1 / NIdToDistH[x])
calculation = float((1 / (sizeGraph - 1)) * hashTableCount)
harmonicList.append([node.GetId(), calculation])
return harmonicList
def predictLinksNegatedShortestPath(GCombined, nodesAtHop, itemNodeIds, userNodeIds, directory):
scores = {}
for node1 in userNodeIds:
for node2 in itemNodeIds:
if not GCombined.IsNode(node1) or not GCombined.IsNode(node2) or GCombined.IsEdge(node1, node2):
if not node1 in scores:
scores[node1] = {}
scores[node1][node2] = 0.0
else:
if not node1 in scores:
scores[node1] = {}
scores[node1][node2] = 1.0/snap.GetShortPath(GCombined, node1, node2, False)
with open(directory + 'NegatedShortestPath', 'wb') as outfile:
pickle.dump(scores, outfile)
def path_proba(graph, name, n=1000):
"""Calculate the probability that a path exists between two uniformly random nodes (n simulations)"""
p = 0
for i in range(n):
a = graph.GetRndNId()
b = graph.GetRndNId()
while a == b:
b = graph.GetRndNId()
NIdToDistH = snap.TIntH()
snap.GetShortPath(graph, a, NIdToDistH, True)
if b in NIdToDistH:
p += 1
print 'Using {} random pairs, the probability that a path exists between two nodes is ' \
'{} for the {} network'.format(n, p / n, name)
return p/n
def getClosenessCentralities(self):
centralities = []
#print("In closeness method")
for NI in self.network.Nodes():
#print("Selected new origin node.\n")
sumShortestPaths = 0
for NI2 in self.network.Nodes():
#print("Selected new comparitive node\n")
sumShortestPaths += abs(snap.GetShortPath(self.network, NI.GetId(), NI2.GetId()))
closeness = float(sumShortestPaths) / float(self.network.GetNodes())
centralities.append(closeness)
#print("Finished getting centrality values\n")
return centralities
def get_dist_distribution(filename, sample_count):
distance_dst = collections.defaultdict(int)
graph = snap.LoadEdgeList(snap.PUNGraph, filename)
node_list = []
for node in graph.Nodes():
node_list.append(node.GetId())
for i in range(0, sample_count):
sample_pair = random.sample(node_list, 2)
dist = snap.GetShortPath(graph, sample_pair[0], sample_pair[1], False)
if dist > 0:
distance_dst[dist] += 1
print "spid is " + str(calculate_spid(distance_dst)) + " for " + str(
sample_count) + " samples"
for item in distance_dst:
distance_dst[item] /= float(sample_count)
return distance_dst
def ssspfun(R0):
paths = []
start0 = time.time()
for i, r in enumerate(R0):
if i % 100 == 0:
text_to_append = "%2.2f percent done, node %i out of %i" % (
100 * double(i) / len(R0), i, len(R0))
print "%2.2f percent done, node %i out of %i" % (
100 * double(i) / len(R0), i, len(R0))
print " %2.2f seconds have elapsed so far..." % (time.time() -
start0)
os.system("echo '" + text_to_append +
"\n' >> percent_done_4testing.txt")
sp = snap.GetShortPath(Graph, r[0], r[1])
paths.append(sp)
print "Done!"
paths = array(paths)
return paths
def get_land_D_mtx(land_ids_temp):
print "Getting landmark -> node distance matrix..."
L2n = zeros((len(land_ids_temp), V0)).astype('int8')
for i, l in enumerate(land_ids_temp):
text_to_append = " getting sssp for node %i out of %i" % (
i + 1, len(land_ids_temp))
print " getting sssp for node %i out of %i" % (i + 1,
len(land_ids_temp))
os.system("echo '" + text_to_append +
"\n' >> percent_done_4landmarks.txt")
sys.stdout.flush()
NIdToDistH = snap.TIntH()
shortestPath = snap.GetShortPath(Graph, int(l), NIdToDistH)
for item in NIdToDistH:
# print item
L2n[i, item - 1] = NIdToDistH[item]
L2n = L2n[:, conn_node_ids - 1]
return L2n
def GetTrustList(fSourceNode, graph, visitPath, totalNodeList, nodeMapNumDict):
# trustList=[0] * len(visitPath)
# tSourceIndex=totalNodeList.index(fSourceNode)
trustList = []
newVisitPath = []
tSourceIndex = nodeMapNumDict[fSourceNode]
# tVisitPathIndex=[totalNodeList.index(node) for node in visitPath]
tVisitPathIndex = [nodeMapNumDict[node] for node in visitPath]
for i in range(len(visitPath)):
node = visitPath[i]
tempDstNodeIndex = nodeMapNumDict[node]
dist = snap.GetShortPath(graph, tSourceIndex, tempDstNodeIndex)
if dist == 1:
newVisitPath.append(node)
trustList.append('1')
return newVisitPath, trustList
def sample_shortest_path(self, n_node=100, isDir=False):
'''
sample diameter, e.g. ‘shortest path’, of a Graph
:param n_node: number of nodes to sample
:param isDir: consider direct or not
'''
snap = self.snap
n_node = min(self.num_nodes, n_node)
nodes = self.nodes
src = np.random.choice(nodes, n_node, replace=False)
dest = np.random.choice(nodes, n_node, replace=False)
ret = []
for i in range(n_node):
Length = snap.GetShortPath(self.graph, int(src[i]), int(dest[i]))
ret.append(Length)
return ret
def q2_4_utils(dataset_name):
G = load_graph(dataset_name)
Rnd = snap.TRnd(42)
Rnd.Randomize()
count = 0
positive_count = 0
negative_count = 0
while count < 1000:
NId_src = G.GetRndNId(Rnd)
NId_dst = G.GetRndNId(Rnd)
if NId_src != NId_dst:
if snap.GetShortPath(G, NId_src, NId_dst, True) > 0:
positive_count = positive_count + 1
else:
negative_count = negative_count + 1
count = count + 1
# print (snap.GetShortPath(G, NId_src, NId_dst))
print 'positive_count', positive_count
print 'negative_count', negative_count