def calculate_stats_users(pct_train):
dat_file = 'user_data_working.csv'
data = Data()
data.load(dat_file,
sep=',',
format={
'col': 0,
'row': 1,
'value': 2,
'ids': int
})
train, test = data.split_train_test(percent=pct_train)
svd = SVD()
svd.set_data(train)
svd.compute(k=100,
min_values=2,
pre_normalize=None,
mean_center=True,
post_normalize=False)
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s\n' % mae.compute()
def ex1(dat_file='./ml-1m/ratings.dat',
pct_train=0.5):
data = Data()
data.load(dat_file, sep='::', format={'col':0, 'row':1, 'value':2,'ids':int})
# create train/test split
train, test = data.split_train_test(percent=pct_train)
# create svd
K=100
svd = SVD()
svd.set_data(train)
svd.compute(k=K, min_values=5, pre_normalize=None, mean_center=True, post_normalize=True)
# evaluate performance
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
def evaluate(data, count=5, K=100):
results = []
for i in range(count):
train, test = data.split_train_test(percent=PERCENT_TRAIN)
print len(data.get()), len(train.get()), len(test.get())
#test_in_train(test, train)
#print train.get()
svd = SVD()
svd.set_data(train)
svd.compute(k=K, min_values=5, pre_normalize=None, mean_center=True, post_normalize=True)
#Evaluation using prediction-based metrics
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
#print "keyerror: ===========================================================>"
continue
try:
rsu = {}
rsu["RMSE"] = rmse.compute()
rsu["MAE"] = mae.compute()
print rsu
results.append(rsu)
except:
print "one error....++++++++++++++++++++++++++++++++++++++++++++++++++++"
return results
def test_SVD(svd,train,test,pct_train):
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s\n' % mae.compute()
def test_SVD(svd, train, test, pct_train):
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s\n' % mae.compute()
def eval_rmse(self):
# Evaluation using prediction-based metrics
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in self.test.get():
try:
pred_rating = self.svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
def eval_reco(model, test):
""" Compute RMSE and MAE on test set
"""
#Evaluation using prediction-based metrics
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = model.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
return rmse, mae
def eval_reco(model, test):
""" Compute RMSE and MAE on test set
"""
#Evaluation using prediction-based metrics
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = model.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
return rmse, mae
def evaluate(_svd, _testData, verbose=False):
global rmse, mae, rating, item_id, user_id, pred_rating
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in _testData.get():
try:
pred_rating = _svd.predict(item_id, user_id, MIN_VALUE=0, MAX_VALUE=10)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
if verbose:
print item_id, user_id, rating, pred_rating
except Exception as e:
print 'ERROR occurred:', e.message
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
def ex1(dat_file=DATA_DIR + 'ml-1m-ratings.dat', pct_train=0.5):
data = Data()
data.load(dat_file,
sep='::',
format={
'col': 0,
'row': 1,
'value': 2,
'ids': int
})
# About format parameter:
# 'row': 1 -> Rows in matrix come from column 1 in ratings.dat file
# 'col': 0 -> Cols in matrix come from column 0 in ratings.dat file
# 'value': 2 -> Values (Mij) in matrix come from column 2 in ratings.dat
# file
# 'ids': int -> Ids (row and col ids) are integers (not strings)
# create train/test split
train, test = data.split_train_test(percent=pct_train)
# create svd
K = 100
svd = SVD()
svd.set_data(train)
svd.compute(k=K,
min_values=5,
pre_normalize=None,
mean_center=True,
post_normalize=True)
# evaluate performance
rmse = RMSE()
# mae is mean ABSOLUTE error
# ... in this case it will return 1.09 which means there is an error of almost 1 point out of 5
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
def evaulte(train_set, test_set):
svd = SVD()
svd.set_data(train_set)
svd.compute(k=KKK, min_values=MIN_ITEM, pre_normalize=None, mean_center=True, post_normalize=True)
mae = MAE()
k_err = 0
for rating, item_id, user_id in test_set.get():
try:
pred_rating = svd.predict(item_id, user_id)
mae.add(rating, pred_rating)
except KeyError:
#print "keyerror: ===========================================================>"
k_err += 1
continue
print "k_err", k_err, " -- ", "test-len: ", len(test_set.get()), "train-len: ", len(train_set.get())
result = mae.compute()/2.0
return result
def calculate_stats_users(pct_train):
dat_file = 'user_data_working.csv'
data = Data()
data.load(dat_file, sep=',', format={'col':0, 'row':1, 'value':2,'ids':int})
train, test = data.split_train_test(percent=pct_train)
svd = SVD()
svd.set_data(train)
svd.compute(k=100, min_values=2, pre_normalize=None, mean_center=True,
post_normalize=False)
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s\n' % mae.compute()
def evaluate(clf, _testData, verbose = False):
rmse = RMSE()
mae = MAE()
numErrors = 0
for rating, item_id, user_id in _testData.get():
try:
pred_rating = clf.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
if verbose:
print item_id, user_id, rating, pred_rating
except KeyError as e:
if verbose:
print 'ERROR occurred:', e.message
numErrors += 1
print '\n%i/%i data points raised errors.' % (numErrors, len(_testData))
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
def evaulte(train_set, test_set):
svd = SVD()
svd.set_data(train_set)
svd.compute(k=KKK,
min_values=MIN_ITEM,
pre_normalize=None,
mean_center=True,
post_normalize=True)
mae = MAE()
k_err = 0
for rating, item_id, user_id in test_set.get():
try:
pred_rating = svd.predict(item_id, user_id)
mae.add(rating, pred_rating)
except KeyError:
#print "keyerror: ===========================================================>"
k_err += 1
continue
print "k_err", k_err, " -- ", "test-len: ", len(
test_set.get()), "train-len: ", len(train_set.get())
result = mae.compute() / 2.0
return result
def ex1(dat_file='ml-1m/ratings.dat',
pct_train=0.5):
data = Data()
data.load(dat_file, sep='::', format={'col':0, 'row':1, 'value':2,
'ids':int})
# About format parameter:
# 'row': 1 -> Rows in matrix come from column 1 in ratings.dat file
# 'col': 0 -> Cols in matrix come from column 0 in ratings.dat file
# 'value': 2 -> Values (Mij) in matrix come from column 2 in ratings.dat
# file
# 'ids': int -> Ids (row and col ids) are integers (not strings)
# create train/test split
train, test = data.split_train_test(percent=pct_train)
# create svd
K = 100
svd = SVD()
svd.set_data(train)
svd.compute(
k=K, min_values=5, pre_normalize=None, mean_center=True, post_normalize=True)
# evaluate performance
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
class TestPrediction(Test):
def __init__(self):
super(TestPrediction, self).__init__()
# Prediction-based metrics: MAE, RMSE, Pearson
self.mae = MAE(self.DATA_PRED)
self.rmse = RMSE(self.DATA_PRED)
self.R = 3 # Real Rating (ground truth)
self.R_PRED = 2.1 # Predicted Rating
# test_PRED MAE
def test_PRED_MAE_compute_one(self):
assert_equal(self.mae.compute(self.R, self.R_PRED), 0.9)
def test_PRED_MAE_compute_one_empty_datasets(self):
mae = MAE()
assert_equal(mae.compute(self.R, self.R_PRED), 0.9)
def test_PRED_MAE_compute_all(self):
assert_equal(self.mae.compute(), 0.7)
def test_PRED_MAE_nan(self):
mae = MAE()
mae.add(2.0, nan)
assert_equal(mae.get_test(), [])
assert_equal(mae.get_ground_truth(), [])
def test_PRED_MAE_load(self):
mae = MAE()
mae.load(self.GT_DATA, self.TEST_DATA)
assert_equal(mae.compute(), 0.7)
def test_PRED_MAE_load_test(self):
mae = MAE()
mae.load_test(self.TEST_DATA)
assert_equal(len(mae.get_test()), len(self.TEST_DATA))
assert_equal(len(mae.get_ground_truth()), 0)
assert_raises(ValueError, mae.compute) #Raise: GT is empty!
def test_PRED_MAE_load_test_and_ground_truth(self):
mae = MAE()
mae.load_test(self.TEST_DATA)
mae.load_ground_truth(self.GT_DATA)
assert_equal(mae.compute(), 0.7)
def test_PRED_MAE_add_entry(self):
self.mae.add(1, 4) #1: GT rating, 4: Predicted rating
assert_equal(len(self.mae.get_test()), len(self.DATA_PRED)+1)
assert_equal(self.mae.compute(), 1.083333)
def test_PRED_MAE_different_list_sizes(self):
mae = MAE()
GT = [3, 1, 5, 2]
# GT list has one element less than self.TEST_DATA
mae.load(GT, self.TEST_DATA)
assert_raises(ValueError, mae.compute)
# test_PRED RMSE
def test_PRED_RMSE_compute_one(self):
#Even though rmse has data, we only compute these two param values
assert_equal(self.rmse.compute(self.R, self.R_PRED), 0.9)
def test_PRED_RMSE_compute_one_empty_datasets(self):
rmse = RMSE()
assert_equal(rmse.compute(self.R, self.R_PRED), 0.9)
def test_PRED_RMSE_compute_all(self):
assert_equal(self.rmse.compute(), 0.891067)
def test_PRED_RMSE_load_test(self):
rmse = RMSE()
self.TEST_DATA = [2.3, 0.9, 4.9, 0.9, 1.5]
rmse.load_test(self.TEST_DATA)
assert_equal(len(rmse.get_test()), len(self.TEST_DATA))
def test_PRED_RMSE_add_entry(self):
self.rmse.add(1,4)
assert_equal(len(self.rmse.get_test()), len(self.DATA_PRED)+1)
assert_equal(self.rmse.compute(), 1.470261)
def test_PRED_RMSE_different_list_sizes(self):
rmse = RMSE()
GT = [3, 1, 5, 2]
# GT list has one element less than self.TEST_DATA
rmse.load(GT, self.TEST_DATA)
assert_raises(ValueError, rmse.compute)
def test_PRED_RMSE_numpy_array(self):
rmse = RMSE()
rmse.load(array(self.GT_DATA), array(self.TEST_DATA))
assert(rmse.compute(), 0.891067)
#Dataset
PERCENT_TRAIN = int(sys.argv[2])
data = Data()
data.load(sys.argv[1], sep='::', format={'col':0, 'row':1, 'value':2, 'ids':int})
#Train & Test data
train, test = data.split_train_test(percent=PERCENT_TRAIN)
svdlibc = SVDLIBC('./ml-1m/ratings.dat')
svdlibc.to_sparse_matrix(sep='::', format={'col':0, 'row':1, 'value':2, 'ids': int})
svdlibc.compute(k=100)
svd = svdlibc.export()
svd.save_model('/tmp/svd-model', options={'k': 100})
#svd.similar(ITEMID1) # results might be different than example 4. as there's no min_values=10 set here
#Evaluation using prediction-based metrics
print 'Evaluating...'
rmse = RMSE()
mae = MAE()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id, 0.0, 5.0)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
except KeyError:
continue
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
post_normalize=True)
# Evaluate
rmse_svd = RMSE()
mae_svd = MAE()
rmse_svd_neig = RMSE()
mae_svd_neig = MAE()
i = 1
total = len(test.get())
print 'Total Test ratings: %s' % total
for rating, item_id, user_id in test:
try:
pred_rating_svd = svd.predict(item_id, user_id)
rmse_svd.add(rating, pred_rating_svd)
mae_svd.add(rating, pred_rating_svd)
pred_rating_svd_neig = svd_neig.predict(item_id,
user_id) #Koren & co.
if pred_rating_svd_neig is not nan:
rmse_svd_neig.add(rating, pred_rating_svd_neig)
mae_svd_neig.add(rating, pred_rating_svd_neig)
print "\rProcessed test rating %d" % i,
sys.stdout.flush()
i += 1
except KeyError:
continue
rmse_svd_all.append(rmse_svd.compute())
def test_PRED_MAE_nan(self):
mae = MAE()
mae.add(2.0, nan)
assert_equal(mae.get_test(), [])
assert_equal(mae.get_ground_truth(), [])
class TestPrediction(Test):
def __init__(self):
super(TestPrediction, self).__init__()
# Prediction-based metrics: MAE, RMSE, Pearson
self.mae = MAE(self.DATA_PRED)
self.rmse = RMSE(self.DATA_PRED)
self.R = 3 # Real Rating (ground truth)
self.R_PRED = 2.1 # Predicted Rating
# test_PRED MAE
def test_PRED_MAE_compute_one(self):
assert_equal(self.mae.compute(self.R, self.R_PRED), 0.9)
def test_PRED_MAE_compute_one_empty_datasets(self):
mae = MAE()
assert_equal(mae.compute(self.R, self.R_PRED), 0.9)
def test_PRED_MAE_compute_all(self):
assert_equal(self.mae.compute(), 0.7)
def test_PRED_MAE_nan(self):
mae = MAE()
mae.add(2.0, nan)
assert_equal(mae.get_test(), [])
assert_equal(mae.get_ground_truth(), [])
def test_PRED_MAE_load(self):
mae = MAE()
mae.load(self.GT_DATA, self.TEST_DATA)
assert_equal(mae.compute(), 0.7)
def test_PRED_MAE_load_test(self):
mae = MAE()
mae.load_test(self.TEST_DATA)
assert_equal(len(mae.get_test()), len(self.TEST_DATA))
assert_equal(len(mae.get_ground_truth()), 0)
assert_raises(ValueError, mae.compute) #Raise: GT is empty!
def test_PRED_MAE_load_test_and_ground_truth(self):
mae = MAE()
mae.load_test(self.TEST_DATA)
mae.load_ground_truth(self.GT_DATA)
assert_equal(mae.compute(), 0.7)
def test_PRED_MAE_add_entry(self):
self.mae.add(1, 4) #1: GT rating, 4: Predicted rating
assert_equal(len(self.mae.get_test()), len(self.DATA_PRED) + 1)
assert_equal(self.mae.compute(), 1.083333)
def test_PRED_MAE_different_list_sizes(self):
mae = MAE()
GT = [3, 1, 5, 2]
# GT list has one element less than self.TEST_DATA
mae.load(GT, self.TEST_DATA)
assert_raises(ValueError, mae.compute)
# test_PRED RMSE
def test_PRED_RMSE_compute_one(self):
#Even though rmse has data, we only compute these two param values
assert_equal(self.rmse.compute(self.R, self.R_PRED), 0.9)
def test_PRED_RMSE_compute_one_empty_datasets(self):
rmse = RMSE()
assert_equal(rmse.compute(self.R, self.R_PRED), 0.9)
def test_PRED_RMSE_compute_all(self):
assert_equal(self.rmse.compute(), 0.891067)
def test_PRED_RMSE_load_test(self):
rmse = RMSE()
self.TEST_DATA = [2.3, 0.9, 4.9, 0.9, 1.5]
rmse.load_test(self.TEST_DATA)
assert_equal(len(rmse.get_test()), len(self.TEST_DATA))
def test_PRED_RMSE_add_entry(self):
self.rmse.add(1, 4)
assert_equal(len(self.rmse.get_test()), len(self.DATA_PRED) + 1)
assert_equal(self.rmse.compute(), 1.470261)
def test_PRED_RMSE_different_list_sizes(self):
rmse = RMSE()
GT = [3, 1, 5, 2]
# GT list has one element less than self.TEST_DATA
rmse.load(GT, self.TEST_DATA)
assert_raises(ValueError, rmse.compute)
def test_PRED_RMSE_numpy_array(self):
rmse = RMSE()
rmse.load(array(self.GT_DATA), array(self.TEST_DATA))
assert (rmse.compute(), 0.891067)
def test_PRED_MAE_nan(self):
mae = MAE()
mae.add(2.0, nan)
assert_equal(mae.get_test(), [])
assert_equal(mae.get_ground_truth(), [])
#Load SVD from /tmp
svd2 = SVD(filename='/tmp/movielens') # Loading already computed SVD model
#Predict User rating for given user and movie:
USERID = 2
ITEMID= 1 # Toy Story
rating1=svd2.predict(ITEMID, USERID, 0.0, 5.0)
print 'Predicted rating=%f'% rating1
flag=0
#Retrieve actual rating for given user and movie
for rating, item_id, user_id in data.get():
if user_id == USERID and item_id == ITEMID:
rat = rating
#print 'Actual rating=%f' % rating
flag=1
break
if flag == 1:
print 'Actual rating=%f'% rat
else :
sys.exit("No actual rating available")
#Evaluating prediction
rmse = RMSE()
mae = MAE()
rmse.add(rating1, rat)
mae.add(rating1, rat)
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
svd_neig.compute(k=K, min_values=None, pre_normalize=None, mean_center=True, post_normalize=True)
# Evaluate
rmse_svd = RMSE()
mae_svd = MAE()
rmse_svd_neig = RMSE()
mae_svd_neig = MAE()
i = 1
total = len(test.get())
print "Total Test ratings: %s" % total
for rating, item_id, user_id in test:
try:
pred_rating_svd = svd.predict(item_id, user_id)
rmse_svd.add(rating, pred_rating_svd)
mae_svd.add(rating, pred_rating_svd)
pred_rating_svd_neig = svd_neig.predict(item_id, user_id) # Koren & co.
if pred_rating_svd_neig is not nan:
rmse_svd_neig.add(rating, pred_rating_svd_neig)
mae_svd_neig.add(rating, pred_rating_svd_neig)
print "\rProcessed test rating %d" % i,
sys.stdout.flush()
i += 1
except KeyError:
continue
rmse_svd_all.append(rmse_svd.compute())
mae_svd_all.append(mae_svd.compute())
MAX_RATING) # predicted rating value
print svd.get_matrix().value(ITEMID, USERID) # real rating value
print ''
print 'GENERATING RECOMMENDATION'
print svd.recommend(USERID, n=5, only_unknowns=True, is_row=False)
#Evaluation using prediction-based metrics
rmse = RMSE()
mae = MAE()
spearman = SpearmanRho()
kendall = KendallTau()
#decision = PrecisionRecallF1()
for rating, item_id, user_id in test.get():
try:
pred_rating = svd.predict(item_id, user_id)
rmse.add(rating, pred_rating)
mae.add(rating, pred_rating)
spearman.add(rating, pred_rating)
kendall.add(rating, pred_rating)
except KeyError:
continue
print ''
print 'EVALUATION RESULT'
print 'RMSE=%s' % rmse.compute()
print 'MAE=%s' % mae.compute()
print 'Spearman\'s rho=%s' % spearman.compute()
print 'Kendall-tau=%s' % kendall.compute()
#print decision.compute()
print ''
