def setUp(self):
self.tf = TensorCoFi(dim=2)
self.df = pd.read_csv(resource_filename(testfm.__name__,
'data/movielenshead.dat'),
sep="::", header=None, names=['user', 'item',
'rating', 'date',
'title'])
self.df = self.df.head(n=100)
def test_score(self):
tf = TensorCoFi(dim=2)
inp = [{'user': 10, 'item': 100}, {'user': 10,'item': 110},
{'user': 12,'item': 120}]
inp = pd.DataFrame(inp)
tf.fit(inp)
uid = tf._dmap['user'][10]
iid = tf._dmap['item'][100]
self.assertEquals(uid, 1)
self.assertEquals(iid, 1)
tf.factors['user'][0][0] = 0
tf.factors['user'][0][1] = 1
tf.factors['item'][0][0] = 1
tf.factors['item'][0][1] = 5
self.assertEqual(0*1+1*5, tf.getScore(10,100))
from testfm.evaluation.parameterTuning import ParameterTuning
eval = Evaluator() # call this before loading the data to save memory (fork of process takes place)
# prepare the data
df = pd.read_csv(
resource_filename(testfm.__name__, "data/movielenshead.dat"),
sep="::",
header=None,
names=["user", "item", "rating", "date", "title"],
)
print df.head()
training, testing = testfm.split.holdoutByRandom(df, 0.9)
print "Tuning the parameters."
tr, validation = testfm.split.holdoutByRandom(training, 0.7)
pt = ParameterTuning()
pt.setMaxIterations(10)
pt.setZvalue(80)
tf_params = pt.getBestParams(TensorCoFi, tr, validation)
print tf_params
tf = TensorCoFi()
tf.setParams(**tf_params)
tf.fit(training)
print tf.getName().ljust(50),
print eval.evaluate_model(tf, testing, all_items=training.item.unique())
eval.close() # need this call to clean up the worker processes
class TestTensorCoFi(unittest.TestCase):
def tearDown(self):
import os
if os.path.exists('user.csv'):
os.remove('user.csv')
if os.path.exists('item.csv'):
os.remove('item.csv')
def setUp(self):
self.tf = TensorCoFi(dim=2)
self.df = pd.read_csv(resource_filename(testfm.__name__,
'data/movielenshead.dat'),
sep="::", header=None, names=['user', 'item',
'rating', 'date',
'title'])
self.df = self.df.head(n=100)
def test_array(self):
arr, tmap = self.tf._dataframe_to_float_matrix(self.df)
def test_fit(self):
self.tf.fit(self.df)
#item and user are row vectors
self.assertEqual(len(self.df.user.unique()), self.tf.factors['user'].shape[0])
self.assertEqual(len(self.df.item.unique()), self.tf.factors['item'].shape[0])
self.assertEqual(self.tf.user_features[1], [1])
self.assertEqual(self.tf.item_features[122], [1])
def test_ids_returns(self):
inp = [{'user': 10, 'item': 100}, {'user': 10, 'item': 110},
{'user': 12,'item': 120}]
inp = pd.DataFrame(inp)
self.tf.fit(inp)
self.assertEquals(self.tf.user_column_names, ['user'])
self.assertEquals(self.tf.item_column_names, ['item'])
self.assertEquals(len(self.tf.user_features), 2)
self.assertEquals(len(self.tf.item_features), 3)
self.assertEquals(len(self.tf.factors['user']), 2)
self.assertEquals(len(self.tf.factors['item']), 3)
uid = self.tf._dmap['user'][10]
iid = self.tf._dmap['item'][100]
self.assertEquals(uid, 1)
self.assertEquals(iid, 1)
self.assertEquals(len(self.tf.factors['user'][uid]), 2)
self.assertEquals(len(self.tf.factors['user'][uid]), self.tf._dim)
self.assertEquals(len(self.tf.factors['item'][iid]), self.tf._dim)
def test_score(self):
tf = TensorCoFi(dim=2)
inp = [{'user': 10, 'item': 100}, {'user': 10,'item': 110},
{'user': 12,'item': 120}]
inp = pd.DataFrame(inp)
tf.fit(inp)
uid = tf._dmap['user'][10]
iid = tf._dmap['item'][100]
self.assertEquals(uid, 1)
self.assertEquals(iid, 1)
tf.factors['user'][0][0] = 0
tf.factors['user'][0][1] = 1
tf.factors['item'][0][0] = 1
tf.factors['item'][0][1] = 5
self.assertEqual(0*1+1*5, tf.getScore(10,100))
def test_floatmatrix_to_numpy(self):
from jnius import autoclass
FloatMatrix = autoclass('org.jblas.FloatMatrix')
rand = FloatMatrix.rand(4,2)
rand_np = self.tf._float_matrix2numpy(rand)
self.assertEqual(rand.get(0,0), rand_np[0,0])
self.assertEqual(rand.get(1,0), rand_np[1,0])
self.assertEqual(rand.get(1,1), rand_np[1,1])
self.assertEqual((rand.rows, rand.columns), rand_np.shape)
def test_score_tcff(self):
tf = TensorCoFiByFile(dim=2)
inp = [{'user': 10, 'item': 100},
{'user': 10,'item': 110},
{'user': 12,'item': 120}]
inp = pd.DataFrame(inp)
tf.fit(inp)
uid = tf._dmap['user'][10]
iid = tf._dmap['item'][100]
self.assertEquals(uid, 1)
self.assertEquals(iid, 1)
tf.factors['user'][0][0] = 0
tf.factors['user'][0][1] = 1
tf.factors['item'][0][0] = 1
tf.factors['item'][0][1] = 5
self.assertEqual(0*1+1*5, tf.getScore(10, 100))
def test_result_by_file(self):
def floatMatrixToCSV(fm, n):
csv = '\n'.join((','.join((str(fm.get(row, column))
for column in xrange(0, fm.columns)))
for row in xrange(0, fm.rows)))
with open(n, 'w') as f:
f.write(csv)
tf = TensorCoFiByFile(dim=2)
inp = [{'user': 10, 'item': 100},
{'user': 10,'item': 110},
{'user': 12,'item': 120}]
inp = pd.DataFrame(inp)
ten = tf._fit(inp)
#####
floatMatrixToCSV(ten.getModel().get(0), 'user.csv')
floatMatrixToCSV(ten.getModel().get(1), 'item.csv')
fromFile = {
'user': np.ma.column_stack(np.genfromtxt(open('user.csv', 'r'),
delimiter=',')),
'item': np.ma.column_stack(np.genfromtxt(open('item.csv', 'r'),
delimiter=','))
}
for i in xrange(0, 2):
self.assertAlmostEqual(tf.factors['user'][i][0],
fromFile['user'][i][0])
self.assertAlmostEqual(tf.factors['user'][i][1],
fromFile['user'][i][1])
for i in xrange(0, 3):
self.assertAlmostEqual(tf.factors['item'][i][0],
fromFile['item'][i][0])
self.assertAlmostEqual(tf.factors['item'][i][1],
fromFile['item'][i][1])