GGGG

createG.py

from snap import *
import csv
import collections
#helpers
def getNeighbours(G, i):     
    neighborsV = TIntV()
    GetNodesAtHop(G, i, 1, neighborsV, False)
    return [i for i in neighborsV]

def getNeighboursOfNeigbours(G, i):     
    neighborsV = TIntV()
    GetNodesAtHop(G, i, 2, neighborsV, False)
    return [i for i in neighborsV]

def isFriend(G, i, j):
	if G.GetIntAttrDatN(i, "person") and G.GetIntAttrDatN(j, "person"):
		return G.IsEdge(i,j)
	return False

def friendCriteria(i, j):
	if i==0: return 0
	ratio = float(i)/j
	if 	 ratio > 0.30: return 10
	elif ratio > 0.27: return 9
	elif ratio > 0.24: return 8
	elif ratio > 0.21: return 7
	elif ratio > 0.18: return 6
	elif ratio > 0.15: return 5
	elif ratio > 0.12: return 4
	elif ratio > 0.09: return 3
	elif ratio > 0.06: return 2
	elif ratio > 0.03: return 1
	else: 			   return 0

def herdingCriteria(i, j):
	if i==0: return 0
	ratio = float(i)/j
	if 	 ratio > 0.60: return 10
	elif ratio > 0.54: return 9
	elif ratio > 0.48: return 8
	elif ratio > 0.42: return 7
	elif ratio > 0.36: return 6
	elif ratio > 0.30: return 5
	elif ratio > 0.24: return 4
	elif ratio > 0.18: return 3
	elif ratio > 0.12: return 2
	elif ratio > 0.06: return 1
	else: 			   return 0

############


# FILENAME = "FoursquareCheckins20110101-20110731.csv"
# FILENAME =  "FScheckinstest.csv" # testni z manj zapis
FILENAME =  "custom.csv" # testni

#Generate GRAPH
def generate(FILENAME):
	G = TNEANet.New()
	Gf = TNEANet.New()

	#read file
	inputfile = csv.reader( open(FILENAME), dialect='excel' )
	inputfile.next() # skip first line

	most_popular_loc = [0, 1]
	#add nodes
	for i in inputfile:
		#avoid 0 and -1
		person = int(i[0]) +1
		location = int(i[4]) * -1 -2

		#add missing nodes
		if not G.IsNode(person):
			G.AddNode(person)
			G.AddIntAttrDatN(person, 1, "person") # 1  = true 0 = False
			Gf.AddNode(person)
			Gf.AddIntAttrDatN(person, 1, "person") # 1  = true 0 = False
		if not G.IsNode(location):
			G.AddNode(location)
			G.AddIntAttrDatN(location, 0, "person")
			Gf.AddNode(location)
			Gf.AddIntAttrDatN(location, 0, "person")

		#to get value from node:
		# val = TStrV()
		# G.StrAttrValueNI(NodeId, val)
		# val[0]
		#add connections
		if not G.IsEdge(person, location):
			EId = G.AddEdge(person, location)
			G.AddIntAttrDatE(EId, 1, "checkin")
			EId = G.AddEdge(location, person)
			G.AddIntAttrDatE(EId, 1, "checkin")
		else:
			EId1 = G.GetEId(person, location)
			EId2 = G.GetEId(location, person)
			EVal = G.GetIntAttrDatE(EId, "checkin")+1

			G.AddIntAttrDatE(EId1, EVal, "checkin") #increase value of checkin if person is twice on same place
			G.AddIntAttrDatE(EId2, EVal, "checkin") #increase value of checkin if person is twice on same place

			#update most popular location
			outDeg = G.GetNI(location).GetOutDeg()
			if most_popular_loc[1] < outDeg:
				most_popular_loc = [location, outDeg]

	#add friends
	# FRIENDSFILE =  "FoursquareFriendship.csv"
	FRIENDSFILE =  "customFriends.csv"
	inputfile = csv.reader( open(FRIENDSFILE), dialect='excel' )
	inputfile.next()

	for i in inputfile:
		if not G.IsNode(int(i[0])+1):
			G.AddNode(int(i[0])+1)
			Gf.AddNode(int(i[0])+1)
		if not G.IsNode( int(i[1])+1 ):
			G.AddNode(int(i[1])+1)
			Gf.AddNode(int(i[1])+1)
		G.AddEdge(int(i[0])+1, int(i[1])+1)
		G.AddEdge(int(i[1])+1, int(i[0])+1)

	#SaveEdgeList(G, "FSGraph.g")
	FOut = TFOut(FILENAME + '.bin')
	G.Save(FOut)
	FOut.Flush()

	FOut = TFOut(FILENAME + 'f.bin')
	Gf.Save(FOut)
	FOut.Flush()

	return most_popular_loc

print generate(FILENAME)

# most_popular_loc = [-81, 33] #location, max checkins
most_popular_loc = [0, 1] #location, max checkins

FIn = TFIn(FILENAME + '.bin')
G = TNEANet.Load(FIn)

FInf = TFIn(FILENAME + 'f.bin')
Gf = TNEANet.Load(FInf)

#G = LoadEdgeList(PNEANet, "FSGraph.g", 0, 1)

print "stevilo vozlisc", G.GetNodes(), "edgov", G.GetEdges()

#-za prijateljstvo..ce je blo na neki lokaciji 30% ali vec prijateljev neke osebe, dobi povezava med to osebo in lokacijo vseh 10 tock
#..27% 9, 24% 8, ... 3% 1 tocko (oz. za eno tocko lah recemo ze, da je bil vsaj kaksen prijatelj tam)
# possibleFriendConnections = collections.defaultdict(lambda: [])

nodes = [i.GetId() for i in G.Nodes()]
persons = [i for i in nodes if G.GetIntAttrDatN(i, "person")]
locations = list(set(nodes) - set(persons))

print "friendships"
for nodeid in persons:

	neigbours = getNeighbours(G, nodeid) #get all connections
	# if G.GetIntAttrDatN(nodeid, "person"): #if node is person
		
	friends = filter(lambda x: G.GetIntAttrDatN(x, "person"), neigbours) #get friends of user
	# places = list(set(neigbours) - set(friends)) #get places where user have checked in

	neighboursOfNeigbours = getNeighboursOfNeigbours(G, nodeid)
	possible_new_places = filter(lambda x: not G.GetIntAttrDatN(x, "person"), neighboursOfNeigbours) #get places of user friends, so at second hop
	
				
	# print "nid", nodeid,"friends",friends,"places",places,"neighboursOfNeigbours", neighboursOfNeigbours, "possible nw places", possible_new_places
	#iterate over possible new places
	for i in possible_new_places:
		
		#count friends on possible new place
		friendCheckin = len([1 for j in friends if G.IsEdge(i, j)])

		#calculate score for new place and save user => [place, score from 0 to 10]
		placeScore = friendCriteria(friendCheckin, len(friends))

		#save score to new edge
		if placeScore > 0:
			EId1 = Gf.AddEdge(nodeid, i)
			EId2 = Gf.AddEdge(i, nodeid)
			Gf.AddIntAttrDatE(EId1, placeScore, "checkin")
			Gf.AddIntAttrDatE(EId2, placeScore, "checkin")
			
			#resitev na key value
			# possibleFriendConnections[nodeid].append( [i, placeScore] )

print "herding"
#herding	
for i in locations:
	neigbours = getNeighbours(G, i) #get all connections
	score = herdingCriteria(len(neigbours), most_popular_loc[1])

	if score > 0: #if > add connection from place to new user
		for j in persons:
			if not G.IsEdge(i, j):
				EId1 = Gf.AddEdge(i, j)
				EId2 = Gf.AddEdge(j, i)

				Gf.AddIntAttrDatE(EId1, placeScore, "herd")
				Gf.AddIntAttrDatE(EId2, placeScore, "herd")



#print possibleFriendConnections.values()