def test_scaleX(self):
c = get_preproc_config(use_exog=True)
d = prepare_data(c)
# train
self.assertEqual(0, int(d.trainX[0][0]))
self.assertEqual(1, int(d.trainX[-1][1]))
def test_input_and_output_shapes(self):
c = get_preproc_config(deseason=False)
df = get_df3()
d = data.Data2d(df, c)
y2 = d._deseasonalize(df['dep_var'], 4)
self.assertEqual(df['dep_var'].shape, y2.shape)
def test_2d_feature_importances_(self):
pc = get_preproc_config(lags=3, horizon=1)
d = prepare_data(pc)
mockConfigSVR = get_mock_svr(d, pc)
r = run_config([d, mockConfigSVR, 'val'])
self.assertEqual(len(r.feature_scores), 3)
self.assertEqual(r.feature_scores[0], ('lag1', 2))
def test_scaleX(self):
c = get_preproc_config(use_exog=True)
d = prepare_data(c)
d = prepare_data(c, dim="3d")
# train
self.assertAlmostEqual(0.0, d.trainX[0][0][0], 2)
self.assertAlmostEqual(1.0, d.trainX[-1][-1][-1], 2)
def test_scaleYref(self):
# assert original Y's are not changed after scaling
c = get_preproc_config(use_exog=True, horizon=1)
d = prepare_data(c)
self.assertEqual(52.0, d.trainYref[0])
self.assertEqual(242.0, d.trainYref[-1])
self.assertEqual(292.0, d.valYref[0])
self.assertEqual(322.0, d.valYref[-1])
self.assertEqual(372.0, d.testYref[0])
self.assertEqual(402.0, d.testYref[-1])
def test_get_data_with_exog__intent2(self):
c = get_preproc_config(lags=3, use_exog=True, intent_distance=2)
d = prepare_data(c)
# 12, 22, 32 are endogs, 2 and 21 are exogs at distance=2
self.assertEqual(d.trainX[0].tolist(), [12, 22, 32, 20, 21])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(), [252, 262, 272, 260, 261])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(), [332, 342, 352, 340, 341])
self.assertEqual(d.testY[0], 362)
def test_get_data_with_exog__intent2(self):
c = get_preproc_config(lags=3, use_exog=True, intent_distance=2)
d = prepare_data(c, dim="3d")
self.assertEqual(d.trainX[0].tolist(),
[[10, 11, 32], [20, 21, 42], [30, 31, 52]])
self.assertEqual(d.trainY[0], 62)
self.assertEqual(d.valX[0].tolist(),
[[250, 251, 272], [260, 261, 282], [270, 271, 292]])
self.assertEqual(d.valY[0], 302)
self.assertEqual(d.testX[0].tolist(),
[[330, 331, 352], [340, 341, 362], [350, 351, 372]])
self.assertEqual(d.testY[0], 382)
def test_deseasonalized_reference_values(self):
c = get_preproc_config(deseason=True,
seasonal_period=4,
horizon=1,
lags=1)
df = get_df3()
d = data.Data2d(df, c)
# all deseasonalized values equal to 10.0
self.assertTrue(np.array_equal(d.trainYref, [10.0] * 11))
self.assertTrue(np.array_equal(d.valYref, [10.0] * 3))
self.assertTrue(np.array_equal(d.testYref, [10.0] * 3))
def get_forecast(self, config, config_dict):
pc = get_preproc_config(lags=3,
use_exog=True,
horizon=1,
feature_selection=0.5,
rfe_step=1,
random_state=1)
d = prepare_data(pc)
c = config(config_dict, pc)
model = c.train(d)
yhat = c.forecast(model, np.array([d.testX[0]]))
return yhat
def test_get_data__no_exog(self):
c = get_preproc_config(lags=3)
d = prepare_data(c)
self.assertEqual(d.trainX.shape, (21, 3))
self.assertEqual(d.valX.shape, (5, 3))
self.assertEqual(d.testX.shape, (5, 3))
self.assertEqual(d.trainX[0].tolist(), [12, 22, 32])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(), [252, 262, 272])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(), [332, 342, 352])
self.assertEqual(d.testY[0], 362)
def test_get_data_with_exog__intent0(self):
c = get_preproc_config(lags=3, use_exog=True)
d = prepare_data(c)
self.assertEqual(d.trainX.shape, (21, 5))
self.assertEqual(d.valX.shape, (5, 5))
self.assertEqual(d.testX.shape, (5, 5))
self.assertEqual(d.trainX[0].tolist(), [12, 22, 32, 40, 41])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(), [252, 262, 272, 280, 281])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(), [332, 342, 352, 360, 361])
self.assertEqual(d.testY[0], 362)
def test_sets(self):
c = get_preproc_config(lags=3)
d_ml = prepare_data(c)
d_lstm = prepare_data(c, dim="3d")
# same shapes
self.assertEqual(d_ml.trainY.shape[0], d_lstm.trainY.shape[0])
self.assertEqual(d_ml.valY.shape[0], d_lstm.valY.shape[0])
self.assertEqual(d_ml.testY.shape[0], d_lstm.testY.shape[0])
# same values
self.assertEqual(d_ml.trainY[0], d_lstm.trainY[0])
self.assertEqual(d_ml.valY[0], d_lstm.valY[0])
self.assertEqual(d_ml.testY[0], d_lstm.testY[0])
def test_get_data__no_exog(self):
c = get_preproc_config(lags=3)
d = prepare_data(c, dim="3d")
self.assertEqual(d.trainX.shape, (21, 3, 1))
self.assertEqual(d.valX.shape, (5, 3, 1))
self.assertEqual(d.testX.shape, (5, 3, 1))
self.assertEqual(d.trainX[0].tolist(), [[12], [22], [32]])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(), [[252], [262], [272]])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(), [[332], [342], [352]])
self.assertEqual(d.testY[0], 362)
def test_scoring(self):
c = get_preproc_config(lags=2,
use_exog=True,
feature_selection=0.5,
random_state=1)
d = prepare_data(c)
# ensure that the lags + most informative exogs are kept
self.assertEqual(tuple(d.trainX[0].tolist()), (100, 20, 20, 100))
# ensure relevant feature_names is left
self.assertEqual(tuple(d.feature_names),
('lag2', 'lag1', 'dim1_0', 'dim1_1'))
def test_select_features(self):
c = get_preproc_config(lags=2,
use_exog=True,
feature_selection=0.5,
random_state=1)
d = prepare_data(c)
# keeping 4 features after selection: 2 original lags and 2 exogs
self.assertEqual(d.trainX.shape[1], 4)
self.assertEqual(d.valX.shape[1], 4)
self.assertEqual(d.testX.shape[1], 4)
# ensure that feature_names is updated too
self.assertEqual(len(d.feature_names), 4)
def test_3d_test_mode(self):
pc = get_preproc_config(lags=3, horizon=1)
d = prepare_data(pc, dim="3d")
mockConfigLSTM = get_mock_lstm(d, pc)
r = run_config([d, mockConfigLSTM, 'test'])
self.assertTrue(r.train_mse > 0.0)
self.assertTrue(r.test_mse > 0.0)
self.assertTrue(r.train_mae > 0.0)
self.assertTrue(r.train_mae > 0.0)
# yhat are scaled
self.assertTrue(r.yhat_is[0] <= 1.0)
self.assertTrue(r.yhat_oos[0] <= 1.0)
def test_2d_val_mode(self):
pc = get_preproc_config(lags=3, horizon=1)
d = prepare_data(pc)
mockConfigSVR = get_mock_svr(d, pc)
r = run_config([d, mockConfigSVR, 'val'])
self.assertTrue(r.train_mse > 0.0)
self.assertTrue(r.test_mse > 0.0)
self.assertTrue(r.train_mae > 0.0)
self.assertTrue(r.train_mae > 0.0)
# yhat are scaled
self.assertTrue(r.yhat_is[0] <= 1.0)
self.assertTrue(r.yhat_oos[0] <= 1.0)
def test_get_data_with_exog__intent0(self):
c = get_preproc_config(lags=3, use_exog=True)
d = prepare_data(c, dim="3d")
self.assertEqual(d.trainX.shape, (21, 3, 3))
self.assertEqual(d.valX.shape, (5, 3, 3))
self.assertEqual(d.testX.shape, (5, 3, 3))
self.assertEqual(d.trainX[0].tolist(),
[[10, 11, 12], [20, 21, 22], [30, 31, 32]])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(),
[[250, 251, 252], [260, 261, 262], [270, 271, 272]])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(),
[[330, 331, 332], [340, 341, 342], [350, 351, 352]])
self.assertEqual(d.testY[0], 362)
def test_get_data_with_exog__all_exog_lags(self):
c = get_preproc_config(lags=3, use_exog=True)
d = prepare_data(c)
self.assertEqual(d.trainX.shape, (21, 9))
self.assertEqual(d.valX.shape, (5, 9))
self.assertEqual(d.testX.shape, (5, 9))
self.assertEqual(d.trainX[0].tolist(),
[12, 22, 32, 10, 11, 20, 21, 30, 31])
self.assertEqual(d.trainY[0], 42)
self.assertEqual(d.valX[0].tolist(),
[252, 262, 272, 250, 251, 260, 261, 270, 271])
self.assertEqual(d.valY[0], 282)
self.assertEqual(d.testX[0].tolist(),
[332, 342, 352, 330, 331, 340, 341, 350, 351])
self.assertEqual(d.testY[0], 362)
def test_scaleY(self):
# ensure scaling works
c = get_preproc_config(use_exog=True, horizon=1)
d = prepare_data(c)
# train
self.assertEqual(0.0, d.trainY[0])
self.assertEqual(1.0, d.trainY[-1])
# val
self.assertTrue(d.valY[c['horizon'] - 1] != d.valYref[0])
self.assertTrue(d.valY[-1] != d.valYref[-1])
# test
self.assertTrue(d.testY[c['horizon'] - 1] != d.testYref[0])
self.assertTrue(d.testY[-1] != d.testYref[-1])
def test_3d_val_mode(self):
pc = get_preproc_config(lags=3, horizon=1)
d = prepare_data(pc, dim="3d")
mockConfigLSTM = get_mock_lstm(d, pc)
print(mockConfigLSTM.__class__.__dict__.keys())
r = run_config([d, mockConfigLSTM, 'val'])
self.assertTrue(r.train_mse > 0.0)
self.assertTrue(r.test_mse > 0.0)
self.assertTrue(r.train_mae > 0.0)
self.assertTrue(r.train_mae > 0.0)
# yhat are scaled
self.assertTrue(r.yhat_is[0] <= 1.0)
self.assertTrue(r.yhat_oos[0] <= 1.0)
def test_horizon(self):
model = Mock()
model.predict = Mock(return_value=[1.])
testX = np.array([
[1, 2, 3, 4, 5, 20, 21],
[2, 3, 4, 5, 6, 20, 22],
[3, 4, 5, 6, 7, 20, 23],
[4, 5, 6, 7, 8, 20, 24],
[5, 6, 7, 8, 9, 20, 25],
[6, 7, 8, 9, 10, 20, 26]
])
pc = get_preproc_config(lags=3, horizon=5)
c = ConfigLSVR({'c': 1., 'eps': 1.}, pc)
yhat = c.forecast(model, testX)
self.assertEqual(yhat.tolist(), [[1], [1]])
def test_instances(self):
"""Ensure predictions are inserted in the right places for
new instances
"""
model = Mock()
model.predict = Mock(return_value=[[1.0]])
# the first col is exogenous vars, the last col is endogenous
testX = np.array([
[[0.0, 0.1, 1.5], [0.0, 0.2, 1.6]],
[[0.0, 0.2, 1.6], [0.0, 0.3, 1.7]],
[[0.0, 0.3, 1.7], [0.0, 0.4, 1.8]],
[[0.0, 0.4, 1.8], [0.0, 0.5, 1.9]],
[[0.0, 0.5, 1.9], [0.0, 0.6, 2.0]],
[[0.0, 0.6, 2.0], [0.0, 0.7, 2.1]],
])
pc = get_preproc_config(lags=2, horizon=2)
adict= {"bidirectional": False, "topology": [3, 4], "epochs": 100,
"batch_size": 10, "activation": None, "dropout_rate": 0.,
"optimizer": None, "kernel_regularizer": (0.0, 0.0),
"bias_regularization": (0.0, 0.0), "early_stopping": None,
"stateful": False}
c = ConfigLSTM(adict, pc)
yhat = c.forecast(model, testX)
calls = model.predict.call_args_list
self.assertEqual(get_call_args(calls[0][0]), [[0., 0.1, 1.5], [0., 0.2, 1.6]])
self.assertEqual(get_call_args(calls[1][0]), [[0., 0.2, 1.6], [0., 0.3, 1.]])
self.assertEqual(get_call_args(calls[2][0]), [[0., 0.2, 1.6], [0., 0.3, 1.7]])
self.assertEqual(get_call_args(calls[3][0]), [[0., 0.3, 1.7], [0., 0.4, 1.]])
self.assertEqual(get_call_args(calls[4][0]), [[0., 0.3, 1.7], [0., 0.4, 1.8]])
self.assertEqual(get_call_args(calls[5][0]), [[0., 0.4, 1.8], [0., 0.5, 1.]])
self.assertEqual(get_call_args(calls[6][0]), [[0., 0.4, 1.8], [0., 0.5, 1.9]])
self.assertEqual(get_call_args(calls[7][0]), [[0., 0.5, 1.9], [0., 0.6, 1.]])
self.assertEqual(get_call_args(calls[8][0]), [[0., 0.5, 1.9], [0., 0.6, 2.]])
self.assertEqual(get_call_args(calls[9][0]), [[0., 0.6, 2.], [0., 0.7, 1.]])
def test_general_case(self):
model = Mock()
model.predict = Mock(return_value=[[1.0]])
testX = np.array([
[[0.0, 0.1, 1.5], [0.0, 0.2, 1.6]],
[[0.0, 0.2, 1.6], [0.0, 0.3, 1.7]],
[[0.0, 0.3, 1.7], [0.0, 0.4, 1.8]],
[[0.0, 0.4, 1.8], [0.0, 0.5, 1.9]],
[[0.0, 0.5, 1.9], [0.0, 0.6, 2.0]],
[[0.0, 0.6, 2.0], [0.0, 0.7, 2.1]]
])
pc = get_preproc_config(lags=2, horizon=1)
adict= {"bidirectional": False, "topology": [3, 4], "epochs": 100,
"batch_size": 10, "activation": None, "dropout_rate": 0.,
"optimizer": None, "kernel_regularizer": (0.0, 0.0),
"bias_regularization": (0.0, 0.0), "early_stopping": None,
"stateful": False}
c = ConfigLSTM(adict, pc)
yhat = c.forecast(model, testX)
self.assertEqual(yhat.tolist(), [[1], [1], [1], [1], [1], [1]])
def test_instances(self):
"""Ensure predictions are inserted in the right places for
new instances
"""
model = Mock()
model.predict = Mock(return_value=[1.])
# the first 3 cols are endogenous vars (from earlier to later),
# the last col is exogenous
testX = np.array([
[.1, .2, .3, 1.5],
[.2, .3, .4, 1.6],
[.3, .4, .5, 1.7],
[.4, .5, .6, 1.8],
[.5, .6, .7, 1.9],
[.6, .7, .8, 2.],
])
pc = get_preproc_config(lags=3, horizon=2, use_exog=True)
c = ConfigLSVR({'c': 1., 'eps': 1.}, pc)
yhat = c.forecast(model, testX)
calls = model.predict.call_args_list
self.assertEqual(get_call_args(calls[0][0]), [0.1, 0.2, 0.3, 1.5])
self.assertEqual(get_call_args(calls[1][0]), [0.2, 0.3, 1, 1.6])
self.assertEqual(get_call_args(calls[2][0]), [0.2, 0.3, 0.4, 1.6])
self.assertEqual(get_call_args(calls[3][0]), [0.3, 0.4, 1, 1.7])
self.assertEqual(get_call_args(calls[4][0]), [0.3, 0.4, 0.5, 1.7])
self.assertEqual(get_call_args(calls[5][0]), [0.4, 0.5, 1, 1.8])
self.assertEqual(get_call_args(calls[6][0]), [0.4, 0.5, 0.6, 1.8])
self.assertEqual(get_call_args(calls[7][0]), [0.5, 0.6, 1, 1.9])
self.assertEqual(get_call_args(calls[8][0]), [0.5, 0.6, 0.7, 1.9])
self.assertEqual(get_call_args(calls[9][0]), [0.6, 0.7, 1, 2.0])
def test_Yref(self):
c = get_preproc_config(lags=3, horizon=1)
d = prepare_data(c, dim="3d")
self.assertEqual(d.trainY[c['horizon'] - 1], d.trainYref[0])
self.assertEqual(d.valY[c['horizon'] - 1], d.valYref[0])
self.assertEqual(d.testY[c['horizon'] - 1], d.testYref[0])
def test_scale_revert(self):
c = get_preproc_config(use_exog=True)
d = prepare_data(c)
d = prepare_data(c, dim="3d")
self.assertTrue(d.revert(d.trainY)[5], d.trainYref[5])
def test_scale_revert(self):
c = get_preproc_config(use_exog=True)
d = prepare_data(c)
self.assertAlmostEqual(d.revert(d.trainY)[5 + c['horizon'] - 1],
d.trainYref[5],
delta=0.9)
