First Version

Author: febrianimanda

jester-data-1.xls

@@ -0,0 +1,101 @@
+__author__ = 'Febrian Imanda Effendy'
+
+import xlrd
+import numpy as np
+import math
+
+# sh = data.sheet_by_index(0)
+# print sh.name, sh.nrows, sh.ncols
+# for rx in range(sh.nrows):
+#   print sh.row(rx)
+def getData(filename):
+  data = xlrd.open_workbook(filename)
+  sheet = data.sheet_by_index(0)
+  return sheet
+
+DATA = getData("jester-data-1500.xls")
+SHEET_ROWS = DATA.nrows
+SHEET_COLUMN = DATA.ncols
+
+# Fungsi untuk mendapatkan rating dari 1 item berdasarkan user
+def getRating(user, item):
+	return DATA.row(user)[item].value
+
+# Fungsi untuk mendapatkan rating dari seluruh item berdasarkan user
+def getItemRating(user):
+	rating = []
+	for item in range(1, SHEET_COLUMN):
+		rating += [getRating(user, item)]
+	# return rating
+	listRating = np.array(rating)
+	return listRating
+
+# Fungsi untuk menghitung rata-rata dari list rating yang diberikan (numpy format)
+def getAverageRating(rates):
+	total = []
+	for i in rates:
+		temp = 0 if i >= 99 else i
+		total.append(temp)
+	listTotal = np.array(total)
+	return np.mean(listTotal)
+
+# Fungsi untuk mendapatkan semua neighbour dari user
+def getNeighbours(user):
+	neighbour = []
+	for i in range(SHEET_ROWS):
+		if i != user :
+			for j in range(SHEET_COLUMN):
+				yUser = getRating(user, j)
+				yNeighbour = getRating(i, j)
+				if (yUser < 99) and (yNeighbour < 99) :
+					neighbour += [i]
+					break
+	# return neighbour
+	listNeighbours = np.array(neighbour)
+	return listNeighbours
+
+# Fungsi untuk mendapatkan similiaritas dari 2 user yang dibandingkan
+def getSimiliarity(user1, user2):
+	yAvgUser1 = getAverageRating(getItemRating(user1))
+	yAvgUser2 = getAverageRating(getItemRating(user2))
+	atas = 0
+	bawah = 0
+	for i in range(SHEET_COLUMN) :
+		yUser1 = getRating(user1, i)
+		yUser2 = getRating(user2, i)
+		atas += (yUser1 - yAvgUser1) * (yUser2 - yAvgUser2)
+	yUser1a = 0
+	yUser2a = 0
+	for i in range(SHEET_COLUMN) :
+		yUser1a += (yUser1 - yAvgUser1) ** 2
+		yUser2a += (yUser2 - yAvgUser2) ** 2
+	bawah = math.sqrt(yUser1a * yUser2a)
+	sim = atas / bawah
+	return sim
+
+# Fungsi untuk mendapatkan 20 similiaritas terbesar menggunakan metode mergesort dengan O(nlog(n))
+def getTopSimiliarity(listSim):
+	listSim = np.sort(listSim, kind='mergesort')
+	listSim = listSim[::-1]
+	return listSim[1:21:1]
+
+# Fungsi untuk mendapatkan prediksi rating
+def getPredictedRating(user, item):
+	yAvgUser = getAverageRating(getItemRating(user))
+	neighbours = getNeighbours(user)
+	atas = 0
+	bawah = 0
+	# for i in range(SHEET_COLUMN):
+ 	for j in range(len(neighbours)):
+ 		similiarities = getSimiliarity(neighbours[j], user)
+ 		tempRating = getRating(neighbours[j], item)
+ 		rating = 0 if tempRating >= 99 else tempRating
+ 		yAvgNeighbour = getAverageRating(getItemRating(neighbours[j]))
+ 		print "User",user, " | User",neighbours[j], " - Similiarities :", similiarities, " - rating :", rating, " - avg :", yAvgNeighbour 
+ 		atas += similiarities * (rating - yAvgNeighbour)
+ 		bawah += abs(similiarities)
+	predicted = yAvgUser + (atas / bawah)
+	return predicted
+
+# print getAverageRating(getItemRating(0))
+print getPredictedRating(0, 100)