def test_memorized_batch_missing():
algo = basic.Memorized(simple_df)
preds = algo.predict(pd.DataFrame({'user': [10, 12, 12], 'item': [1, 1, 3]}))
assert set(preds.index) == set([0, 1, 2])
assert all(preds.iloc[:2] == [4.0, 3.0])
assert np.isnan(preds.iloc[2])
def test_fallback_predict():
algo = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
algo.fit(lktu.ml_pandas.renamed.ratings)
assert len(algo.algorithms) == 2
bias = algo.algorithms[1]
assert isinstance(bias, basic.Bias)
assert bias.mean_ == approx(lktu.ml_pandas.ratings.rating.mean())
# first user + item
preds = algo.predict_for_user(10, [1])
assert preds.loc[1] == 4.0
# second user + first item
preds = algo.predict_for_user(15, [1])
assert preds.loc[1] == approx(bias.mean_ + bias.user_offsets_.loc[15] +
bias.item_offsets_.loc[1])
# second item + user item
preds = algo.predict_for_user(12, [2])
assert preds.loc[2] == approx(bias.mean_ + bias.user_offsets_.loc[12] +
bias.item_offsets_.loc[2])
# blended
preds = algo.predict_for_user(10, [1, 5])
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(bias.mean_ + bias.user_offsets_.loc[10] +
bias.item_offsets_.loc[5])
# blended unknown
preds = algo.predict_for_user(10, [5, 1, -23081])
assert len(preds) == 3
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(bias.mean_ + bias.user_offsets_.loc[10] +
bias.item_offsets_.loc[5])
assert preds.loc[-23081] == approx(bias.mean_ + bias.user_offsets_.loc[10])
def test_memorized_batch_keep_index():
algo = basic.Memorized(simple_df)
query = pd.DataFrame({'user': [10, 10, 12], 'item': [1, 2, 1]},
index=np.random.choice(np.arange(10), 3, False))
preds = algo.predict(query)
assert all(preds.index == query.index)
assert all(preds == [4.0, 5.0, 3.0])
def test_memorized_batch():
algo = basic.Memorized(simple_df)
preds = algo.predict(pd.DataFrame({'user': [10, 10, 12], 'item': [1, 2, 1]}))
assert isinstance(preds, pd.Series)
assert preds.name == 'prediction'
assert set(preds.index) == set([0, 1, 2])
assert all(preds == [4.0, 5.0, 3.0])
def test_fallback_save_load(tmp_path):
tmp_path = lktu.norm_path(tmp_path)
original = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
original.fit(lktu.ml_pandas.renamed.ratings)
fn = tmp_path / 'fallback'
original.save(fn)
algo = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
algo.load(fn)
bias = algo.algorithms[1]
assert bias.mean_ == approx(lktu.ml_pandas.ratings.rating.mean())
def exp_val(user, item):
v = bias.mean_
if user is not None:
v += bias.user_offsets_.loc[user]
if item is not None:
v += bias.item_offsets_.loc[item]
return v
# first user + item
preds = algo.predict_for_user(10, [1])
assert preds.loc[1] == 4.0
# second user + first item
preds = algo.predict_for_user(15, [1])
assert preds.loc[1] == approx(exp_val(15, 1))
# second item + user item
preds = algo.predict_for_user(12, [2])
assert preds.loc[2] == approx(exp_val(12, 2))
# blended
preds = algo.predict_for_user(10, [1, 5])
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
# blended unknown
preds = algo.predict_for_user(10, [5, 1, -23081])
assert len(preds) == 3
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
assert preds.loc[-23081] == approx(exp_val(10, None))
def test_memorized_batch_ord():
algo = basic.Memorized(simple_df)
preds = algo.predict(
pd.DataFrame({
'user': [10, 12, 10],
'item': [1, 1, 2]
}))
assert set(preds.index) == set([0, 1, 2])
assert all(preds == [4.0, 3.0, 5.0])
def test_fallback_clone():
algo = basic.Fallback([basic.Memorized(simple_df), basic.Bias()])
algo.fit(lktu.ml_test.ratings)
assert len(algo.algorithms) == 2
clone = lku.clone(algo)
assert clone is not algo
for a1, a2 in zip(algo.algorithms, clone.algorithms):
assert a1 is not a2
assert type(a2) == type(a1)
def test_fallback_list():
algo = basic.Fallback([basic.Memorized(simple_df), basic.Bias()])
algo.fit(lktu.ml_test.ratings)
assert len(algo.algorithms) == 2
params = algo.get_params()
assert list(params.keys()) == ['algorithms']
assert len(params['algorithms']) == 2
assert isinstance(params['algorithms'][0], basic.Memorized)
assert isinstance(params['algorithms'][1], basic.Bias)
def test_memorized():
algo = basic.Memorized(simple_df)
preds = algo.predict_for_user(10, [1, 2])
assert set(preds.index) == set([1, 2])
assert all(preds == pd.Series({1: 4.0, 2: 5.0}))
preds = algo.predict_for_user(12, [1, 3])
assert set(preds.index) == set([1, 3])
assert preds.loc[1] == 3.0
assert np.isnan(preds.loc[3])
def test_fallback_save_load(tmp_path):
tmp_path = lktu.norm_path(tmp_path)
original = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
original.fit(lktu.ml_pandas.renamed.ratings)
fn = tmp_path / 'fallback'
original.save(fn)
algo = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
algo.load(fn)
bias = algo.algorithms[1]
assert bias.mean_ == approx(lktu.ml_pandas.ratings.rating.mean())
# first user + item
preds = algo.predict_for_user(10, [1])
assert preds.loc[1] == 4.0
# second user + first item
preds = algo.predict_for_user(15, [1])
assert preds.loc[1] == approx(bias.mean_ + bias.user_offsets_.loc[15] +
bias.item_offsets_.loc[1])
# second item + user item
preds = algo.predict_for_user(12, [2])
assert preds.loc[2] == approx(bias.mean_ + bias.user_offsets_.loc[12] +
bias.item_offsets_.loc[2])
# blended
preds = algo.predict_for_user(10, [1, 5])
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(bias.mean_ + bias.user_offsets_.loc[10] +
bias.item_offsets_.loc[5])
# blended unknown
preds = algo.predict_for_user(10, [5, 1, -23081])
assert len(preds) == 3
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(bias.mean_ + bias.user_offsets_.loc[10] +
bias.item_offsets_.loc[5])
assert preds.loc[-23081] == approx(bias.mean_ + bias.user_offsets_.loc[10])
def test_fallback_train_one_pred_impossible():
algo = basic.Fallback(basic.Memorized(simple_df))
algo.fit(lktu.ml_test.ratings)
preds = algo.predict_for_user(10, [1, 2])
assert set(preds.index) == set([1, 2])
assert all(preds == pd.Series({1: 4.0, 2: 5.0}))
preds = algo.predict_for_user(12, [1, 3])
assert set(preds.index) == set([1, 3])
assert preds.loc[1] == 3.0
assert np.isnan(preds.loc[3])
def test_topn_recommend():
pred = basic.Memorized(simple_df)
rec = basic.TopN(pred)
rec10 = rec.recommend(10, candidates=[1, 2])
assert all(rec10.item == [2, 1])
assert all(rec10.score == [5, 4])
rec2 = rec.recommend(12, candidates=[1, 2])
assert len(rec2) == 1
assert all(rec2.item == [1])
assert all(rec2.score == [3])
rec10 = rec.recommend(10, n=1, candidates=[1, 2])
assert len(rec10) == 1
assert all(rec10.item == [2])
assert all(rec10.score == [5])
def test_fallback_save_load(tmp_path):
original = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
original.fit(lktu.ml_test.ratings)
fn = tmp_path / 'fb.mod'
with fn.open('wb') as f:
pickle.dump(original, f)
with fn.open('rb') as f:
algo = pickle.load(f)
bias = algo.algorithms[1]
assert bias.mean_ == approx(lktu.ml_test.ratings.rating.mean())
def exp_val(user, item):
v = bias.mean_
if user is not None:
v += bias.user_offsets_.loc[user]
if item is not None:
v += bias.item_offsets_.loc[item]
return v
# first user + item
preds = algo.predict_for_user(10, [1])
assert preds.loc[1] == 4.0
# second user + first item
preds = algo.predict_for_user(15, [1])
assert preds.loc[1] == approx(exp_val(15, 1))
# second item + user item
preds = algo.predict_for_user(12, [2])
assert preds.loc[2] == approx(exp_val(12, 2))
# blended
preds = algo.predict_for_user(10, [1, 5])
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
# blended unknown
preds = algo.predict_for_user(10, [5, 1, -23081])
assert len(preds) == 3
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
assert preds.loc[-23081] == approx(exp_val(10, None))
def test_fallback_predict():
algo = basic.Fallback(basic.Memorized(simple_df), basic.Bias())
algo.fit(lktu.ml_test.ratings)
assert len(algo.algorithms) == 2
bias = algo.algorithms[1]
assert isinstance(bias, basic.Bias)
assert bias.mean_ == approx(lktu.ml_test.ratings.rating.mean())
def exp_val(user, item):
v = bias.mean_
if user is not None:
v += bias.user_offsets_.loc[user]
if item is not None:
v += bias.item_offsets_.loc[item]
return v
# first user + item
preds = algo.predict_for_user(10, [1])
assert preds.loc[1] == 4.0
# second user + first item
preds = algo.predict_for_user(15, [1])
assert preds.loc[1] == approx(exp_val(15, 1))
# second item + user item
preds = algo.predict_for_user(12, [2])
assert preds.loc[2] == approx(exp_val(12, 2))
# blended
preds = algo.predict_for_user(10, [1, 5])
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
# blended unknown
preds = algo.predict_for_user(10, [5, 1, -23081])
assert len(preds) == 3
assert preds.loc[1] == 4.0
assert preds.loc[5] == approx(exp_val(10, 5))
assert preds.loc[-23081] == approx(exp_val(10, None))
def test_topn_config():
pred = basic.Memorized(simple_df)
rec = basic.TopN(pred)
rs = str(rec)
assert rs.startswith('TopN/')
def test_fallback_string():
algo = basic.Fallback([basic.Memorized(simple_df), basic.Bias()])
assert 'Fallback' in str(algo)