import sys,random, math, operator
from operator import itemgetter
random.seed(0)
class UserBasedCF:
def __init__(self):
self.trainset = {}
self.testset = {}
self.resultset = {}
self.mean = {}
self.user_max = 0
self.item_max = 0
self.sim_vector = {}
self.sim_rank = {}
self.mse = 0.0
@staticmethod
def loadfile(filename):
fp = open(filename, 'r')
for i, line in enumerate(fp):
yield line.strip('\r\n')
fp.close()
def generate_dataset(self, filename, pivot=0.7):
trainset_len = 0
testset_len = 0
for line in UserBasedCF.loadfile(filename):
user, movie, rating, timestamp = line.split('\t')
user = int(user)
movie = int(movie)
rating = int(rating)
self.user_max = max(self.user_max, user)
self.item_max = max(self.item_max, movie)
self.trainset.setdefault(user, {})
self.testset.setdefault(user, {})
if random.random() < pivot:
self.trainset[user][movie] = rating
trainset_len += 1
else:
self.testset[user][movie] = rating
testset_len += 1
for user in range(1, self.user_max+1):
all = 0.0
num = 0
for item in range(1, self.item_max+1):
if item in self.trainset[user]:
all += self.trainset[user][item]
num += 1
self.mean[user] = all / num
print testset_len
def cal_user_sim(self):
self.sim_vector.clear()
for u1 in range(1, self.user_max+1):
self.sim_vector.setdefault(u1, {})
# if u1 % 100 == 0:
# print u1
for u2 in range(u1+1, self.user_max+1):
self.sim_vector.setdefault(u2, {})
xy = 0.0
xx = 0.0
yy = 0.0
if u1 in self.trainset and u2 in self.trainset:
for index in range(1, self.item_max+1):
if index in self.trainset[u1] and index in self.trainset[u2]:
xy += self.trainset[u1][index] * self.trainset[u2][index]
xx += self.trainset[u1][index] * self.trainset[u1][index]
yy += self.trainset[u2][index] * self.trainset[u2][index]
if math.sqrt(xx)*math.sqrt(yy) > 0:
self.sim_vector[u1][u2] = xy / (math.sqrt(xx)*math.sqrt(yy))
self.sim_vector[u2][u1] = self.sim_vector[u1][u2]
else:
self.sim_vector[u1][u2] = 0
self.sim_vector[u2][u1] = self.sim_vector[u1][u2]
def cal_user_sim2(self):
self.sim_vector.clear()
for u1 in range(1, self.user_max+1):
self.sim_vector.setdefault(u1, {})
# if u1 % 100 == 0:
# print u1
for u2 in range(u1+1, self.user_max+1):
self.sim_vector.setdefault(u2, {})
xy = 0.0
xx = 0.0
yy = 0.0
num_both = 0.0
num1 = 0.0
num2 = 0.0
if u1 in self.trainset and u2 in self.trainset:
for i in range(1, self.item_max+1):
if i in self.trainset[u1]:
num1 += 1
if i in self.trainset[u2]:
num2 += 1
if i in self.trainset[u1] and i in self.trainset[u2]:
num_both += 1
xy += self.trainset[u1][i] * self.trainset[u2][i]
xx += self.trainset[u1][i] * self.trainset[u1][i]
yy += self.trainset[u2][i] * self.trainset[u2][i]
if math.sqrt(xx) * math.sqrt(yy) > 0:
self.sim_vector[u1][u2] = xy / (math.sqrt(xx)*math.sqrt(yy)) * num_both * num_both / num1 / num2
self.sim_vector[u2][u1] = self.sim_vector[u1][u2]
else:
self.sim_vector[u1][u2] = 0
self.sim_vector[u2][u1] = self.sim_vector[u1][u2]
def getRecommendation(self, k):
self.sim_rank.clear()
self.resultset.clear()
for i in range(1, self.user_max+1):
self.sim_rank.setdefault(1, {})
self.sim_rank[i] = sorted(self.sim_vector[i].iteritems(), key=lambda x: x[1], reverse=True)
for i in range(1, self.user_max+1):
self.resultset.setdefault(i, {})
for j in range(1, self.item_max+1):
sim_sum = 0.0
if j not in self.trainset[i]:
self.resultset[i][j] = 0.0
for kk in range(0, k):
sim_user = self.sim_rank[i][kk][0]
sim_value = self.sim_rank[i][kk][1]
if j in self.trainset[sim_user]:
self.resultset[i][j] += (self.trainset[sim_user][j] - self.mean[sim_user]) * sim_value
sim_sum += sim_value
self.resultset[i][j] /= sim_sum
self.resultset[i][j] += self.mean[i]
if self.resultset[i][j] < 0:
self.resultset[i][j] = 0
def evaluation(self):
mse_cnt = 0
self.mse = 0.0
for i in range(1, self.user_max+1):
for j in range(1, self.item_max+1):
if j in self.testset[i]:
self.mse += (self.testset[i][j] - self.resultset[i][j]) * (self.testset[i][j] - self.resultset[i][j])
mse_cnt += 1
self.mse /= mse_cnt
if __name__ == '__main__':
usercf = UserBasedCF()
usercf.generate_dataset('u.data')
for i in range(5, 35, 5):
print '---------------------------------------'
usercf.cal_user_sim()
usercf.getRecommendation(i)
usercf.evaluation()
print i, usercf.mse
usercf.cal_user_sim2()
usercf.getRecommendation(i)
usercf.evaluation()
print i, usercf.mse
