def test_local_not_existing_capper_False_estimatePreference(self):
userID = "Leopoldo Pires"
itemID = "You, Me and Dupree"
recSys = UserRecommender(self.model, self.similarity, self.neighbor, False)
self.assertAlmostEquals(
2.065394689, recSys.estimatePreference(userID=userID, similarity=self.similarity, itemID=itemID)
)
def test_local_estimatePreference(self):
userID = "Marcel Caraciolo"
itemID = "Superman Returns"
recSys = UserRecommender(self.model, self.similarity, self.neighbor, True)
self.assertAlmostEquals(
3.5, recSys.estimatePreference(userID=userID, similarity=self.similarity, itemID=itemID)
)
def test_evaluate_RMSRecommenderEvaluator(self):
evaluator = RMSRecommenderEvaluator()
recommender = UserRecommender(self.model, self.similarity,
self.neighbor, True)
evaluationPercentage = 1.0
trainingPercentage = 0.7
numUsers = self.model.NumUsers()
trainingUsers = {}
testUserPrefs = {}
self.total = 0
self.diffs = 0.0
for userID in self.model.UserIDs():
if random() < evaluationPercentage:
evaluator.processOneUser(trainingPercentage, trainingUsers,
testUserPrefs, userID, self.model)
total_training = sum([
len([pref for pref in prefs])
for user, prefs in trainingUsers.iteritems()
])
total_testing = sum([
len([pref for pref in prefs])
for user, prefs in testUserPrefs.iteritems()
])
#self.assertAlmostEquals(total_training/float(total_training+total_testing), 0.7)
#self.assertAlmostEquals(total_testing/float(total_training+total_testing), 0.3)
trainingModel = DictDataModel(trainingUsers)
self.assertEquals(sorted(trainingModel.UserIDs()),
sorted([user for user in trainingUsers]))
recommender.model = trainingModel
self.assertEquals(recommender.model, trainingModel)
for userID, prefs in testUserPrefs.iteritems():
estimatedPreference = None
for pref in prefs:
try:
estimatedPreference = recommender.estimatePreference(
userID=userID, similarity=self.similarity, itemID=pref)
except:
pass
if estimatedPreference is not None:
estimatedPreference = evaluator.capEstimatePreference(
estimatedPreference)
self.assert_(
estimatedPreference <= evaluator.maxPreference
and estimatedPreference >= evaluator.minPreference)
diff = prefs[pref] - estimatedPreference
self.diffs += (diff * diff)
self.total += 1
result = sqrt(self.diffs / float(self.total))
def test_local_not_existing_rescorer_estimatePreference(self):
userID = "Leopoldo Pires"
itemID = "You, Me and Dupree"
recSys = UserRecommender(self.model, self.similarity, self.neighbor, False)
scorer = TanHScorer()
self.assertAlmostEquals(
2.5761016605,
recSys.estimatePreference(userID=userID, similarity=self.similarity, itemID=itemID, rescorer=scorer),
)
def test_local_not_existing_capper_False_estimatePreference(self):
userID = 'Leopoldo Pires'
itemID = 'You, Me and Dupree'
recSys = UserRecommender(self.model, self.similarity, self.neighbor,
False)
self.assertAlmostEquals(
2.065394689,
recSys.estimatePreference(userID=userID,
similarity=self.similarity,
itemID=itemID))
def test_local_estimatePreference(self):
userID = 'Marcel Caraciolo'
itemID = 'Superman Returns'
recSys = UserRecommender(self.model, self.similarity, self.neighbor,
True)
self.assertAlmostEquals(
3.5,
recSys.estimatePreference(userID=userID,
similarity=self.similarity,
itemID=itemID))
def test_local_not_existing_rescorer_estimatePreference(self):
userID = 'Leopoldo Pires'
itemID = 'You, Me and Dupree'
recSys = UserRecommender(self.model, self.similarity, self.neighbor,
False)
scorer = TanHScorer()
self.assertAlmostEquals(
2.5761016605,
recSys.estimatePreference(userID=userID,
similarity=self.similarity,
itemID=itemID,
rescorer=scorer))
def test_evaluate_RMSRecommenderEvaluator(self):
evaluator = RMSRecommenderEvaluator()
recommender = UserRecommender(self.model,self.similarity,self.neighbor,True)
evaluationPercentage = 1.0
trainingPercentage = 0.7
numUsers = self.model.NumUsers()
trainingUsers = {}
testUserPrefs = {}
self.total = 0
self.diffs = 0.0
for userID in self.model.UserIDs():
if random() < evaluationPercentage:
evaluator.processOneUser(trainingPercentage,trainingUsers,testUserPrefs,userID,self.model)
total_training = sum([ len([pref for pref in prefs]) for user,prefs in trainingUsers.iteritems()])
total_testing = sum([ len([pref for pref in prefs]) for user,prefs in testUserPrefs.iteritems()])
#self.assertAlmostEquals(total_training/float(total_training+total_testing), 0.7)
#self.assertAlmostEquals(total_testing/float(total_training+total_testing), 0.3)
trainingModel = DictDataModel(trainingUsers)
self.assertEquals(sorted(trainingModel.UserIDs()), sorted([user for user in trainingUsers]))
recommender.model = trainingModel
self.assertEquals(recommender.model,trainingModel)
for userID,prefs in testUserPrefs.iteritems():
estimatedPreference = None
for pref in prefs:
try:
estimatedPreference = recommender.estimatePreference(userID=userID,similarity=self.similarity,itemID=pref)
except:
pass
if estimatedPreference is not None:
estimatedPreference = evaluator.capEstimatePreference(estimatedPreference)
self.assert_(estimatedPreference <= evaluator.maxPreference and estimatedPreference >= evaluator.minPreference)
diff = prefs[pref] - estimatedPreference
self.diffs+= (diff * diff)
self.total += 1
result = sqrt(self.diffs / float(self.total))
