def test_global_modeling_split_df():
### GLOBAL MODELLING - SPLIT DF
log.info("Global Modeling - Split df")
df = pd.read_csv(PEYTON_FILE, nrows=512)
df1 = df.iloc[:128, :].copy(deep=True)
df2 = df.iloc[128:256, :].copy(deep=True)
df3 = df.iloc[256:384, :].copy(deep=True)
df_dict = {"dataset1": df1, "dataset2": df2, "dataset3": df3}
m = NeuralProphet(n_forecasts=2, n_lags=3)
log.info("split df with single df")
df_train, df_val = m.split_df(df1)
log.info("split df with dict of dataframes")
df_train, df_val = m.split_df(df_dict)
log.info("split df with dict of dataframes - local_split")
df_train, df_val = m.split_df(df_dict, local_split=True)
def test_train_eval_test(self):
log.info("testing: Train Eval Test")
m = NeuralProphet(
n_lags=10,
n_forecasts=3,
ar_sparsity=0.1,
epochs=3,
batch_size=32,
)
df = pd.read_csv(PEYTON_FILE, nrows=95)
df = df_utils.check_dataframe(df, check_y=False)
df = m._handle_missing_data(df, freq="D", predicting=False)
df_train, df_test = m.split_df(df,
freq="D",
valid_p=0.1,
inputs_overbleed=True)
metrics = m.fit(df_train,
freq="D",
validate_each_epoch=True,
valid_p=0.1)
metrics = m.fit(df_train, freq="D")
val_metrics = m.test(df_test)
log.debug("Metrics: train/eval: \n {}".format(
metrics.to_string(float_format=lambda x: "{:6.3f}".format(x))))
log.debug("Metrics: test: \n {}".format(
val_metrics.to_string(float_format=lambda x: "{:6.3f}".format(x))))
def check_split(df_in,
df_len_expected,
n_lags,
n_forecasts,
freq,
p=0.1):
m = NeuralProphet(
n_lags=n_lags,
n_forecasts=n_forecasts,
)
df_in = df_utils.check_dataframe(df_in, check_y=False)
df_in = m._handle_missing_data(df_in, freq=freq, predicting=False)
assert df_len_expected == len(df_in)
total_samples = len(df_in) - n_lags - 2 * n_forecasts + 2
df_train, df_test = m.split_df(df_in,
freq=freq,
valid_p=0.1,
inputs_overbleed=True)
n_train = len(df_train) - n_lags - n_forecasts + 1
n_test = len(df_test) - n_lags - n_forecasts + 1
assert total_samples == n_train + n_test
n_test_expected = max(1, int(total_samples * p))
n_train_expected = total_samples - n_test_expected
assert n_train == n_train_expected
assert n_test == n_test_expected
def test_train_eval_test(self):
log.info("testing: Train Eval Test")
m = NeuralProphet(
n_lags=14,
n_forecasts=7,
ar_sparsity=0.1,
epochs=2,
)
df = pd.read_csv(PEYTON_FILE, nrows=512)
df_train, df_test = m.split_df(df, valid_p=0.1, inputs_overbleed=True)
metrics = m.fit(df_train, freq="D", validate_each_epoch=True, valid_p=0.1)
val_metrics = m.test(df_test)
log.debug("Metrics: train/eval: \n {}".format(metrics.to_string(float_format=lambda x: "{:6.3f}".format(x))))
log.debug("Metrics: test: \n {}".format(val_metrics.to_string(float_format=lambda x: "{:6.3f}".format(x))))
def test_infer_frequency():
df = pd.read_csv(PEYTON_FILE, nrows=102)[:50]
m = NeuralProphet()
# Check if freq is set automatically
df_train, df_test = m.split_df(df)
log.debug("freq automatically set")
# Check if freq is set when equal to the original
df_train, df_test = m.split_df(df, freq="D")
log.debug("freq is equal to ideal")
# Check if freq is set in different freq
df_train, df_test = m.split_df(df, freq="5D")
log.debug("freq is set even though is different than the ideal")
# Assert for data unevenly spaced
index = np.unique(np.geomspace(1, 40, 20, dtype=int))
df_uneven = df.iloc[index, :]
with pytest.raises(ValueError):
m.split_df(df_uneven)
# Check if freq is set even in a df with multiple freqs
df_train, df_test = m.split_df(df_uneven, freq="H")
log.debug("freq is set even with not definable freq")
# Check if freq is set for list
df_dict = {"df1": df, "df2": df}
m = NeuralProphet()
m.fit(df_dict, epochs=5)
log.debug("freq is set for list of dataframes")
# Check if freq is set for list with different freq for n_lags=0
df1 = df.copy(deep=True)
time_range = pd.date_range(start="1994-12-01",
periods=df.shape[0],
freq="M")
df1["ds"] = time_range
df_dict = {"df1": df, "df2": df1}
m = NeuralProphet(n_lags=0, epochs=5)
m.fit(df_dict, epochs=5)
log.debug("freq is set for list of dataframes(n_lags=0)")
# Assert for automatic frequency in list with different freq
m = NeuralProphet(n_lags=2)
with pytest.raises(ValueError):
m.fit(df_dict, epochs=5)
# Exceptions
frequencies = ["M", "MS", "Y", "YS", "Q", "QS", "B", "BH"]
df = df.iloc[:200, :]
for freq in frequencies:
df1 = df.copy(deep=True)
time_range = pd.date_range(start="1994-12-01",
periods=df.shape[0],
freq=freq)
df1["ds"] = time_range
df_train, df_test = m.split_df(df1)
log.debug("freq is set for all the exceptions")
