Greykite: A flexible, intuitive and fast forecasting library

Hi,

Looking at the setup.py file, it looks like the following are all required runtime dependencies, all of which need to be pinned very precisely:

requirements = [    "Cython==0.29.23",    "cvxpy==1.1.12",    "fbprophet==0.5",    "holidays==0.9.10",  # 0.10.2,    "ipykernel==4.8.2",    "ipython==7.1.1",    "ipywidgets==7.2.1",    "jupyter==1.0.0",    "jupyter-client==6.1.5",    "jupyter-console==6.",  # used version 6 to avoid conflict with ipython version    "jupyter-core==4.7.1",    "matplotlib==3.4.1",    "nbformat==5.1.3",    "notebook==5.4.1",    "numpy==1.20.2",    "osqp==0.6.1",    "overrides==2.8.0",    "pandas==1.1.3",    "patsy==0.5.1",    "Pillow==8.0.1",    "plotly==3.10.0",    "pystan==2.18.0.0",    "pyzmq==22.0.3",    "scipy==1.5.4",    "seaborn==0.9.0",    "six==1.15.0",    "scikit-learn==0.24.1",    "Sphinx==3.2.1",    "sphinx-gallery==0.6.1",    "sphinx-rtd-theme==0.4.2",    "statsmodels==0.12.2",    "testfixtures==6.14.2",    "tornado==5.1.1",    "tqdm==4.52.0"]

My question is - why pin them so tightly, and are all of them really necessary? E.g. do I really need sphinx-gallery? Such tight pins make it very difficult to integrate into any existing project. Why not just require a lower bound for many/most of these?

Hi, Thank you for providing this wonderful forecasting package! I've having the best time exploring the greykite package!

However I ran into a tiny issue about regressors:

I'm forecasting with regressor on this type of data: So my regressor is already forecasted and no lagging needed. But I keep getting this error saying "AttributeError: 'dict' object has no attribute 'lagged_regressors'"

FYI This is the config I've been using:

metadata = MetadataParam(
    time_col='date', 
    value_col='fake_target',
    freq='D', 
    train_end_date = '2022-01-10'
)
evaluation_period = EvaluationPeriodParam(
    cv_max_splits=0 # This is to disable CV for demo purposes and just train it on the full data
)
evaluation_metric = EvaluationMetricParam(
    cv_selection_metric=EvaluationMetricEnum.RootMeanSquaredError.name
)

model_components = dict(
    growth=dict(growth_term=None),
    regressors=dict(
        regressor_cols=["fake_regressor"]
    ),
    autoregression=dict(autoreg_dict=None),
    lagged_regressors = dict(lagged_regressor_dict = None) # This is the confusing part, I explicitly said NO lagged terms
)

config = ForecastConfig(
    model_template=ModelTemplateEnum.SILVERKITE.name,
    forecast_horizon=5, 
    coverage=0.95, 
    metadata_param=metadata,
    evaluation_period_param=evaluation_period,
    evaluation_metric_param=evaluation_metric,
    model_components_param=model_components
)

# And run
forecaster = Forecaster()
result = forecaster.run_forecast_config( 
    df=train,
    config=config
)

And here's a screenshot of the error

I was wondering where I did wrong?

Thank you in advance for your support and have a wonderful day! All the best, Kathy Gao

Even basic implementation of greykite (see below) does not pickle properly, due to some of the design choices within Greykite (e.g. nested functions and namedtuple definitions within function class calls.

Was this a purposeful design choice? Is there another method to save a trained model state and reuse the model to create inferences downstream? Integrations with deployment tools become much more challenging if we need to retrain the model every time and can't save the model state. Looking for guidance here on best practice - thanks!

Here's code to reproduce the issue:

from greykite.framework.templates.autogen.forecast_config import ForecastConfig
from greykite.framework.templates.autogen.forecast_config import MetadataParam
from greykite.framework.templates.forecaster import Forecaster
from greykite.framework.templates.model_templates import ModelTemplateEnum

import pandas as pd
import numpy as np

date_list = pd.date_range(start='2020-01-01', end='2022-01-01', freq='W-FRI')
df_train = pd.DataFrame(
    {
        'week_end_date': date_list,
        'data': np.random.rand(len(date_list))
    }
)

metadata = MetadataParam(
    time_col="week_end_date",
    value_col=df_train.columns[-1],
    freq='W-FRI'
)

fc = Forecaster()
result = fc.run_forecast_config(
    df=df_train,
    config=ForecastConfig(
        model_template=ModelTemplateEnum.SILVERKITE.name,
        forecast_horizon=52,
        coverage=0.95,         # 95% prediction intervals
        metadata_param=metadata
    )
)

import dill
with open("pickle_out.b", "wb") as fp:
    dill.dump(result.model, fp)
    output_set = dill.load(fp)

Hi,

First of all thank you for open-sourcing this library. It's really complete and well though (as well as the Silverkite algorithm itself).

However, I think I have spotted a potential bug:

It seems that the option seasonality_changepoints_dict in ModelComponentsParam does seem to break some functionality within pandas, when running Silverkite with cross-validation.

Here's a complete example (using Greykite 0.2.0):

import pandas as pd
import numpy as np

# Load airline passengers dataset (with monthly data):
air_passengers = pd.read_csv("https://raw.githubusercontent.com/jbrownlee/Datasets/master/airline-passengers.csv")
air_passengers["Month"] = pd.to_datetime(air_passengers["Month"])
air_passengers = air_passengers.set_index("Month").asfreq("MS").reset_index()

# Prepare Greykite configs:
from greykite.framework.templates.autogen.forecast_config import (ComputationParam, 
                                                                  EvaluationMetricParam, 
                                                                  EvaluationPeriodParam,
                                                                  ForecastConfig, 
                                                                  MetadataParam, 
                                                                  ModelComponentsParam)

# Metadata:
metadata_params = MetadataParam(date_format=None,  # infer
                                freq="MS",
                                time_col="Month",
                                train_end_date=None,
                                value_col="Passengers")

# Eval metric:
evaluation_metric_params = EvaluationMetricParam(agg_func=np.sum,   # Sum all forecasts...
                                                 agg_periods=12,    # ...Over 12 months
                                                 cv_report_metrics=["MeanSquaredError", "MeanAbsoluteError", "MeanAbsolutePercentError"],
                                                 cv_selection_metric="MeanAbsolutePercentError",
                                                 null_model_params=None,
                                                 relative_error_tolerance=None)

# Eval procedure (CV & backtest):
evaluation_period_params = EvaluationPeriodParam(cv_expanding_window=False,
                                                 cv_horizon=0,   # No CV for now. CHANGE THIS
                                                 cv_max_splits=5,
                                                 cv_min_train_periods=24,
                                                 cv_periods_between_splits=6,
                                                 cv_periods_between_train_test=0,
                                                 cv_use_most_recent_splits=False,
                                                 periods_between_train_test=0,
                                                 test_horizon=12)

# Config for seasonality changepoints
seasonality_components_df = pd.DataFrame({"name": ["conti_year"],
                                          "period": [1.0],
                                          "order": [5],
                                          "seas_names": ["yearly"]})

# Model components (quite long):
model_components_params = ModelComponentsParam(autoregression={"autoreg_dict": "auto"},
                                               
                                               changepoints={"changepoints_dict":  [{"method":"auto",
                                                                                     "potential_changepoint_n": 50,
                                                                                     "no_changepoint_proportion_from_end": 0.2,
                                                                                     "regularization_strength": 0.01}],
                                                             
                                                             # Seasonality changepoints
                                                             "seasonality_changepoints_dict": [{"regularization_strength": 0.6,
                                                                                                "no_changepoint_proportion_from_end": 0.8,
                                                                                                "seasonality_components_df": seasonality_components_df,
                                                                                                "potential_changepoint_n": 50,
                                                                                                "resample_freq":"MS"},
                                                                                               ]
                                                            },
                                               
                                               custom={"fit_algorithm_dict": [{"fit_algorithm": "linear"},
                                                                              ],
                                                       "feature_sets_enabled": "auto",
                                                       "min_admissible_value": 0.0},
                                               
                                               events={"holiday_lookup_countries": None,
                                                       "holidays_to_model_separately": None,
                                                       },
                                               
                                               growth={"growth_term":["linear"]},
                                               
                                               hyperparameter_override={"input__response__outlier__z_cutoff": [100.0],
                                                                        "input__response__null__impute_algorithm": ["ts_interpolate"]},
                                               
                                               regressors=None,
                                               
                                               lagged_regressors=None,
                                               
                                               seasonality={"yearly_seasonality": [5],
                                                            "quarterly_seasonality": ["auto"],
                                                            "monthly_seasonality": False,
                                                            "weekly_seasonality": False,
                                                            "daily_seasonality": False},
                                               
                                               uncertainty=None)

# Computation
computation_params = ComputationParam(n_jobs=1,
                                      verbose=3)


# Define forecaster:
from greykite.framework.templates.forecaster import Forecaster

# defines forecast configuration
config=ForecastConfig(model_template="SILVERKITE",
                      forecast_horizon=12,
                      coverage=0.8,
                      metadata_param=metadata_params,
                      evaluation_metric_param=evaluation_metric_params,
                      evaluation_period_param=evaluation_period_params,
                      model_components_param=model_components_params,
                      computation_param=computation_params,
                     )

# Run:
# creates forecast
forecaster = Forecaster()
result = forecaster.run_forecast_config(df=air_passengers, 
                                        config=config 
                                        )

If we run the piece of code above, everything works as expected. However, if we activate cross-validation (increasing cv_horizon to 5 for instance), Greykite crashes. This happens unless we remove seasonality changepoints (through removing seasonality_changepoints_dict).

The crash traceback looks as follows:

5 fits failed out of a total of 5.
The score on these train-test partitions for these parameters will be set to nan.
If these failures are not expected, you can try to debug them by setting error_score='raise'.

Below are more details about the failures:
--------------------------------------------------------------------------------
5 fits failed with the following error:
Traceback (most recent call last):
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\sklearn\model_selection\_validation.py", line 681, in _fit_and_score
    estimator.fit(X_train, y_train, **fit_params)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\sklearn\pipeline.py", line 394, in fit
    self._final_estimator.fit(Xt, y, **fit_params_last_step)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\sklearn\estimator\simple_silverkite_estimator.py", line 239, in fit
    self.model_dict = self.silverkite.forecast_simple(
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\algo\forecast\silverkite\forecast_simple_silverkite.py", line 708, in forecast_simple
    trained_model = super().forecast(**parameters)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\algo\forecast\silverkite\forecast_silverkite.py", line 719, in forecast
    seasonality_changepoint_result = get_seasonality_changepoints(
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\algo\changepoint\adalasso\changepoint_detector.py", line 1177, in get_seasonality_changepoints
    result = cd.find_seasonality_changepoints(**seasonality_changepoint_detection_args)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\common\python_utils.py", line 787, in fn_ignore
    return fn(*args, **kwargs)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\algo\changepoint\adalasso\changepoint_detector.py", line 736, in find_seasonality_changepoints
    seasonality_df = build_seasonality_feature_df_with_changes(
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\greykite\algo\changepoint\adalasso\changepoints_utils.py", line 237, in build_seasonality_feature_df_with_changes
    fs_truncated_df.loc[(features_df["datetime"] < date).values, cols] = 0
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\indexing.py", line 719, in __setitem__
    indexer = self._get_setitem_indexer(key)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\indexing.py", line 646, in _get_setitem_indexer
    self._ensure_listlike_indexer(key)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\indexing.py", line 709, in _ensure_listlike_indexer
    self.obj._mgr = self.obj._mgr.reindex_axis(
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\internals\base.py", line 89, in reindex_axis
    return self.reindex_indexer(
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\internals\managers.py", line 670, in reindex_indexer
    self.axes[axis]._validate_can_reindex(indexer)
  File "C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\pandas\core\indexes\base.py", line 3785, in _validate_can_reindex
    raise ValueError("cannot reindex from a duplicate axis")
ValueError: cannot reindex from a duplicate axis


C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\sklearn\model_selection\_search.py:969: UserWarning:

One or more of the test scores are non-finite: [nan]

C:\Users\SOTOVJU1\Anaconda3\envs\greykite\lib\site-packages\sklearn\model_selection\_search.py:969: UserWarning:

One or more of the train scores are non-finite: [nan]

It would be great to cross-validate when seasonality changepoint is activated, as it allows to learn multiplicative seasonalities for instance in a similar fashion as Prophet or Orbit do.

Thank you!

Hello all,

I am running the Greykite without cross-validation (cv_max_splits = 0) because I am using the LassoCV() algorithm which by itself uses 5-fold CV. The ForecastConfig() is as follows, in particular, evaluation_metric is all set to None because cv_max_splits = 0:

However, the output on the console suggests that at least 3 metrics are evaluated. My response contains zeros so I do not want MAPE and MedAPE to be reported, and I do not want "Correlation" to be reported either. As a matter of fact, since the loss function in LassoCV() is MSE (L2-norm), I am not interested in anything rather than MSE, really. Unless the loss function in LassoCV() could be changed to MAE (L1-norm) in that case I would be interested in the MAE instead of MSE:

Do you have any suggestions please ?

Best regards, Dario

Hello

I'm working in some house price time series using Greykite but for some reason, the forecast I got is just a median price between upper and lower (ARIMA), so is this known issue with Greykite when we have a pure linear increasing series?

Thank you Aktham Momani

Hi Team,

Can you please confirm if multi-stage forecasting works on weekly aggregation as well.

I tried with data that has daily frequency, so for one stage I kept the daily frequency & next stage its the weekly aggregation of the daily data.

But getting the below error

TypeError: '<' not supported between instances of 'pandas._libs.tslibs.offsets.Day' and 'pandas._libs.tslibs.offsets.Week'

Hi all,

I have added 69 regressors to ModelComponentsParam, so my model instance is as follows: ModelComponentsParam(autoregression={'autoreg_dict': 'auto'}, changepoints={'changepoints_dict': [None, {'method': 'auto'}]}, custom={'fit_algorithm_dict': [{'fit_algorithm': 'elastic_net', 'fit_algorithm_params': {'l1_ratio': array([0.01, 0.1 , 0.2 , 0.3 , 0.4 , 0.5 , 0.6 , 0.7 , 0.8 , 0.9 , 0.99]), 'n_alphas': 100, 'alphas': None, 'fit_intercept': True, 'cv': None, 'tol': 0.001, 'max_iter': 1000}}], 'feature_sets_enabled': 'auto', 'min_admissible_value': None, 'max_admissible_value': None}, events={'holiday_lookup_countries': [], 'holidays_to_model_separately': None, 'holiday_pre_num_days': 2, 'holiday_post_num_days': 2, 'holiday_pre_post_num_dict': None, 'daily_event_df_dict': None}, growth={'growth_term': None}, hyperparameter_override={'input__response__outlier__use_fit_baseline': False, 'input__response__outlier__z_cutoff': None, 'input__response__null__impute_algorithm': None, 'input__regressors_numeric__outlier__use_fit_baseline': False, 'input__regressors_numeric__outlier__z_cutoff': None, 'input__regressors_numeric__null__impute_algorithm': None, 'input__regressors_numeric__normalize__normalize_algorithm': 'RobustScaler', 'input__regressors_numeric__normalize__normalize_params': {'quantile_range': (10.0, 90.0)}, 'degenerate__drop_degenerate': False}, regressors={'regressor_cols': ['ownSame_5_NET_PRICE_BAG', 'media_total_spend', 'Discount_Depth', 'weather_wghtd_avg_tmp_flslk_2m_f_max_low', 'ownSame_3_NET_PRICE_BAG', 'media_tv_traditional', 'hol_EasterSunday', 'cg_STRNGNCY_LGCY_INDX_MEAN', 'hol_RestorationofIndependence_LAG1', 'NET_PRICE_BAG', 'hol_AllSaintsDay_LAG3', 'hol_holiday_count', 'hol_AssumptionDay_lead4', 'hol_AssumptionDay_lead2', 'cc_AVG_IR_MEAN', 'hol_spain_perc_wknd', 'media_digital', 'hol_CorpusChristi_LAG2', 'hol_AssumptionDay_lead3', 'hol_CorpusChristi_LAG3', 'inp_actual_inventory', 'weather_wghtd_avg_tmp_flslk_2m_f_min_high', 'gm_AVG_PRKS_PCNT_CHNG_FRM_BSLNE_MEAN', 'weather_wghtd_avg_tmp_flslk_2m_f_max_high', 'hol_holiday_count_lead4', 'hol_spain_hol_flag', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_mid', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_high', 'hol_ConstitutionDay_lead2', 'hol_portugal_perc_longwknd', 'gm_AVG_RSDNTL_AND_PHRMCY_PCNT_CHNG_FRM_BSLNE_MEAN', 'weather_wghtd_avg_tmp_flslk_2m_f_min_low', 'weather_presence_of_snow', 'hol_ImmaculateConception', 'hol_portugal_fri_mon_flag', 'ownSame_2_NET_PRICE_BAG', 'gm_AVG_RTL_AND_RCRTN_PCNT_CHNG_FRM_BSLNE_MEAN', 'hol_spain_fri_mon_flag', 'hol_NationalDay_LAG4', 'Dollar_Discount', 'hol_portugal_perc_wknd', 'weather_wghtd_avg_tmp_flslk_2m_f_min_mid', 'hol_NationalDay_LAG3', 'ownSame_1_NET_PRICE_BAG', 'hol_portugal_hol_flag', 'cg_STRNGNCY_INDX_MEAN', 'hol_ConstitutionDay', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_low', 'hol_Epiphany_lead4', 'cg_CNTNMT_HLTH_INDX_MEAN', 'cc_AVG_CFR_MEAN', 'hol_spain_perc_longwknd', 'weather_wghtd_avg_tmp_flslk_2m_f_max_mid', 'cg_GOVT_RSPNS_INDX_MEAN', 'inp_projected_inventory', 'ph_Unemployment_persons', 'ownSame_4_NET_PRICE_BAG', 'cc_REC_MEAN', 'inp_actual_inventory_flag', 'hol_NationalDay_LAG2', 'hol_AssumptionDay_lead1', 'hol_GoodFriday', 'hol_CorpusChristi_LAG4', 'hol_spain_wknd_flag', 'hol_portugal_wknd_flag', 'weather_wghtd_avg_cld_cvr_tot_pct_max_mid', 'weather_wghtd_avg_cld_cvr_tot_pct_max_low', 'weather_wghtd_avg_cld_cvr_tot_pct_max_high', 'media_total_spend_lag4']}, lagged_regressors=None, seasonality={'yearly_seasonality': 'auto', 'quarterly_seasonality': 'auto', 'monthly_seasonality': 'auto', 'weekly_seasonality': False, 'daily_seasonality': False}, uncertainty=None)

When checking the model output summary, I only find 65 regressors in the regressor_features category. Of the missing 4, 3 have ended up in the trend_features category, and 1 in the lag_features category:

summary = result.model[-1].summary()
for key, val in summary.pred_category.items():
    print(key)
    print(len(val))
    print(val)

intercept 1 ['Intercept'] time_features 0 [] event_features 0 [] trend_features 3 ['weather_wghtd_avg_cld_cvr_tot_pct_max_mid', 'weather_wghtd_avg_cld_cvr_tot_pct_max_low', 'weather_wghtd_avg_cld_cvr_tot_pct_max_high'] seasonality_features 44 ['sin1_tom_monthly', 'cos1_tom_monthly', 'sin2_tom_monthly', 'cos2_tom_monthly', 'sin1_toq_quarterly', 'cos1_toq_quarterly', 'sin2_toq_quarterly', 'cos2_toq_quarterly', 'sin3_toq_quarterly', 'cos3_toq_quarterly', 'sin4_toq_quarterly', 'cos4_toq_quarterly', 'sin5_toq_quarterly', 'cos5_toq_quarterly', 'sin1_ct1_yearly', 'cos1_ct1_yearly', 'sin2_ct1_yearly', 'cos2_ct1_yearly', 'sin3_ct1_yearly', 'cos3_ct1_yearly', 'sin4_ct1_yearly', 'cos4_ct1_yearly', 'sin5_ct1_yearly', 'cos5_ct1_yearly', 'sin6_ct1_yearly', 'cos6_ct1_yearly', 'sin7_ct1_yearly', 'cos7_ct1_yearly', 'sin8_ct1_yearly', 'cos8_ct1_yearly', 'sin9_ct1_yearly', 'cos9_ct1_yearly', 'sin10_ct1_yearly', 'cos10_ct1_yearly', 'sin11_ct1_yearly', 'cos11_ct1_yearly', 'sin12_ct1_yearly', 'cos12_ct1_yearly', 'sin13_ct1_yearly', 'cos13_ct1_yearly', 'sin14_ct1_yearly', 'cos14_ct1_yearly', 'sin15_ct1_yearly', 'cos15_ct1_yearly'] lag_features 1 ['media_total_spend_lag4'] regressor_features 65 ['ownSame_5_NET_PRICE_BAG', 'media_total_spend', 'Discount_Depth', 'weather_wghtd_avg_tmp_flslk_2m_f_max_low', 'ownSame_3_NET_PRICE_BAG', 'media_tv_traditional', 'hol_EasterSunday', 'cg_STRNGNCY_LGCY_INDX_MEAN', 'hol_RestorationofIndependence_LAG1', 'NET_PRICE_BAG', 'hol_AllSaintsDay_LAG3', 'hol_holiday_count', 'hol_AssumptionDay_lead4', 'hol_AssumptionDay_lead2', 'cc_AVG_IR_MEAN', 'hol_spain_perc_wknd', 'media_digital', 'hol_CorpusChristi_LAG2', 'hol_AssumptionDay_lead3', 'hol_CorpusChristi_LAG3', 'inp_actual_inventory', 'weather_wghtd_avg_tmp_flslk_2m_f_min_high', 'gm_AVG_PRKS_PCNT_CHNG_FRM_BSLNE_MEAN', 'weather_wghtd_avg_tmp_flslk_2m_f_max_high', 'hol_holiday_count_lead4', 'hol_spain_hol_flag', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_mid', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_high', 'hol_ConstitutionDay_lead2', 'hol_portugal_perc_longwknd', 'gm_AVG_RSDNTL_AND_PHRMCY_PCNT_CHNG_FRM_BSLNE_MEAN', 'weather_wghtd_avg_tmp_flslk_2m_f_min_low', 'weather_presence_of_snow', 'hol_ImmaculateConception', 'hol_portugal_fri_mon_flag', 'ownSame_2_NET_PRICE_BAG', 'gm_AVG_RTL_AND_RCRTN_PCNT_CHNG_FRM_BSLNE_MEAN', 'hol_spain_fri_mon_flag', 'hol_NationalDay_LAG4', 'Dollar_Discount', 'hol_portugal_perc_wknd', 'weather_wghtd_avg_tmp_flslk_2m_f_min_mid', 'hol_NationalDay_LAG3', 'ownSame_1_NET_PRICE_BAG', 'hol_portugal_hol_flag', 'cg_STRNGNCY_INDX_MEAN', 'hol_ConstitutionDay', 'weather_wghtd_avg_tmp_flslk_2m_f_mean_low', 'hol_Epiphany_lead4', 'cg_CNTNMT_HLTH_INDX_MEAN', 'cc_AVG_CFR_MEAN', 'hol_spain_perc_longwknd', 'weather_wghtd_avg_tmp_flslk_2m_f_max_mid', 'cg_GOVT_RSPNS_INDX_MEAN', 'inp_projected_inventory', 'ph_Unemployment_persons', 'ownSame_4_NET_PRICE_BAG', 'cc_REC_MEAN', 'inp_actual_inventory_flag', 'hol_NationalDay_LAG2', 'hol_AssumptionDay_lead1', 'hol_GoodFriday', 'hol_CorpusChristi_LAG4', 'hol_spain_wknd_flag', 'hol_portugal_wknd_flag'] interaction_features 0 []

Best, Dario

Hi all,

When setting fit_algorithm_params={"cv": 5} to use 5-fold CV with sklearn LassoCV() on the training set, how should the global parameter "cv_max_splits" be set up ? (either set it to zero, or to None - equivalent to 3 - or equal to 5 ?).

Best regards, Dario

• I'm trying to fit GreyKite Model to my time series data.

• I have attached the csv file for reference.

• Even though the model works, it raises a bunch of warnings that I'd like to avoid.

• Since some of my target values are zero it tells me that MAPE is undefined.

• Also, since I'm only forecasting one step into the future, it gives me an UndefinedMetricWarning : R^2 score is not well-defined with less than two samples.'

• I have attached a few images displaying the warnings.

• Any help to get rid of these warnings would be appreciated!

• This is the code I'm using to fit the data:

`class GreyKiteModel(AnomalyModel):

def __init__(self, *args,model_kwargs = {}, **kwargs) -> None:
    super().__init__(*args, **kwargs)
    self.model_kwargs = model_kwargs
    
def predict(self, df: pd.DataFrame, ) -> pd.DataFrame:
    """Takes in pd.DataFrame with 2 columns, dt and y, and returns a 
    pd.DataFrame with 3 columns, dt, y, and yhat_lower, yhat_upper.

    :param df: Input Dataframe with dt, y columns
    :type df: pd.DataFrame
    :return: Output Dataframe with dt, y, yhat_lower, yhat_upper 
    columns
    :rtype: pd.DataFrame
    """
    df = df.rename(columns = {"dt":"ds", "y":"y"})
    metadata = MetadataParam(time_col="ds", # ----> name of the time column 
                             value_col="y", # ----> name of the value column 
                             freq="D"       # ----> H" for hourly, "D" for daily, "W" for weekly, etc. 
                            )
    forecaster = Forecaster()  # Creates forecasts and stores the result
    result = forecaster.run_forecast_config(df=df, # result is also stored as forecaster.forecast_result
                                            config=ForecastConfig(model_template=ModelTemplateEnum.SILVERKITE.name,
                                                                  forecast_horizon=1,  # forecasts 1 step
                                                                  coverage=0.95,
                                                                  metadata_param=metadata 
                                                                  )
                                            )
    forecast_df = result.forecast.df
    forecast_df = forecast_df.drop(columns=['actual'])
    forecast_df.rename(columns={'ds':'dt',
                                'forecast':'y', 
                                'forecast_lower':'yhat_lower', 
                                'forecast_upper':'yhat_upper' },
                       inplace=True)
    return forecast_df`

df.csv

I'm trying to deploy my greykite model on GCP via a cloud function. The existing read and write functions only work for local directories and not cloud blob storage options. I've adjusted the write function to write to cloud storage but the load function is proving to be a bit challenging.

I am working with people count data, they are clearly positive integers, including 0. How can GreyKite support >=0 data? In other words, is it possible to tell the model that negative values are forbidden by adding a very large penalty in the cost function?

The load_obj in the pickle_utils.py file makes lots of use of this pattern: dill.load(open(os.path.join(dir_name, file), 'rb')) And since dill does not close the file handler passed to the load method, it causes a Resource Warning for each of the parts of the model being loaded. In the following case, it caused 2856 of those warnings flooding the logs.

c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4762240Z 
2022-08-20T16:56:02.4762597Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4.type'>
2022-08-20T16:56:02.4762842Z 
2022-08-20T16:56:02.4763280Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4764328Z 
2022-08-20T16:56:02.4764749Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\df_dropna__value__.type'>
2022-08-20T16:56:02.4765023Z 
2022-08-20T16:56:02.4765471Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4765803Z 
2022-08-20T16:56:02.4766176Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\df__value__.type'>
2022-08-20T16:56:02.4766440Z 
2022-08-20T16:56:02.4766859Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4767206Z 
2022-08-20T16:56:02.4767571Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\features_df__value__.type'>
2022-08-20T16:56:02.4767866Z 
2022-08-20T16:56:02.4768299Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4768626Z 
2022-08-20T16:56:02.4769008Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\fitted_df__value__.type'>
2022-08-20T16:56:02.4769279Z 
2022-08-20T16:56:02.4769714Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4770043Z 
2022-08-20T16:56:02.4770431Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\fs_components_df__value__.type'>
2022-08-20T16:56:02.4770714Z 
2022-08-20T16:56:02.4771148Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4771476Z 
2022-08-20T16:56:02.4771863Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\ml_model_summary__value__.type'>
2022-08-20T16:56:02.4772143Z 
2022-08-20T16:56:02.4772588Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4772917Z 
2022-08-20T16:56:02.4773290Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\uncertainty_model__value__.type'>
2022-08-20T16:56:02.4773589Z 
2022-08-20T16:56:02.4774004Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4774354Z 
2022-08-20T16:56:02.4774721Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\x_design_info__value__.type'>
2022-08-20T16:56:02.4775010Z 
2022-08-20T16:56:02.4775425Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:313: ResourceWarning:
2022-08-20T16:56:02.4776102Z 
2022-08-20T16:56:02.4776491Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\x_mat__value__.type'>
2022-08-20T16:56:02.4776755Z 
2022-08-20T16:56:02.4777188Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:376: ResourceWarning:
2022-08-20T16:56:02.4777512Z 
2022-08-20T16:56:02.4777900Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\adjust_anomalous_info__key__.pkl'>
2022-08-20T16:56:02.4778437Z 
2022-08-20T16:56:02.4778891Z c:\windows\serviceprofiles\networkservice\.conda\envs\aml_test_38\lib\site-packages\greykite\framework\templates\pickle_utils.py:395: ResourceWarning:
2022-08-20T16:56:02.4779216Z 
2022-08-20T16:56:02.4779615Z unclosed file <_io.BufferedReader name='C:\\Windows\\SERVIC~2\\NETWOR~1\\AppData\\Local\\Temp\\tmp55hy74nb\\tmp7q08pus4\\adjust_anomalous_info__value__.pkl'>
...

We were using graykite 0.2.0, but I see the issue is still present in version 0.4.0.

If instead we had a simple read_pickle or dill_load function that would handle closing the file:

def read_pickle(path, mode):
    """Loads the pickled files and closes the file handle.

    Parameters
    ----------
    path : `str`
        The path to the pickled file.
    mode : `str`
        The mode of the open function.
    """
    with open(path, mode) as file:
        data = dill.load(file)
    return data

It would be just a question of changing the calls to dill.load(open(...)) to read_pickle(...). Are there any reason not to use this pattern?

I have some time series forecasts of weekly data where forecasts are too smooth or not really pronounced (not sure how to describe it). Another time series however, for the same dates, same general value range, same model parameters, shows "correct" variability. Both datasets show the same warnings when run (generally about daily event timestamps due to weekly data and a warning that data is highly fragmented)

In another dataset (I have not currently access to), removing the last entry from the time-series fixed the variability. This does however not work for the dataset1 below.

Is this a bug or intended behaviour? If this is intended, which parameters would I have to tweak to mitigate this? Thx for your help

Dataset1: relatively smooth forcast (False)

Dataset2: correct variability forecast (True)

I am using the following settings for the model (Greykite v.0.4.0):

- ForecastConfig(model_template=ModelTemplateEnum.SILVERKITE.name,  forecast_horizon=16, coverage=0.95)
- metadata = MetadataParam(anomaly_info=None, date_format=None, freq='W', time_col='date', train_end_date=None, value_col='value')
- model_components = 
{ 'autoregression': None,
    'changepoints': {
        'changepoints_dict': {
            'method': 'auto',
            'yearly_seasonality_order': 15,
            'regularization_strength': 0.5,
            'resample_freq': '7D',
            'potential_changepoint_n': 25,
            'no_changepoint_proportion_from_end': 0.2
        }
    },
    'custom': {
        'fit_algorithm_dict': {
            'fit_algorithm': 'ridge'
        }
    },
    'events': {
        'holidays_to_model_separately': 'ALL_HOLIDAYS_IN_COUNTRIES',
        'holiday_lookup_countries': ['DE'],
        'holiday_pre_num_days': 4,
        'holiday_post_num_days': 1,
        'holiday_pre_post_num_dict': None,
        'daily_event_df_dict': {
            'Mothersday': date event_name
            0 2017-05-14 Mothersday
            1 2018-05-13 Mothersday
            2 2019-05-12 Mothersday
            3 2020-05-10 Mothersday
            4 2021-05-09 Mothersday
            5 2022-05-08 Mothersday
            6 2023-05-14 Mothersday
            7 2024-05-12 Mothersday
            8 2025-05-11 Mothersday,
            'Black Friday': date event_name
            0 2017-11-24 Black Friday
            1 2018-11-23 Black Friday
            2 2019-11-29 Black Friday
            3 2020-11-27 Black Friday
            4 2021-11-26 Black Friday
            5 2022-11-25 Black Friday
            6 2023-11-24 Black Friday
            7 2024-11-29 Black Friday
            8 2025-11-29 Black Friday
        }
    },
    'growth': None,
    'hyperparameter_override': None,
    'regressors': None,
    'lagged_regressors': None,
    'seasonality': None,
    'uncertainty': None
}

DataSet1 (False):

df1 = pd.DataFrame.from_dict({'date':{0:'2017-12-31',1:'2018-01-07',2:'2018-01-14',3:'2018-01-21',4:'2018-01-28',5:'2018-02-04',6:'2018-02-11',7:'2018-02-18',8:'2018-02-25',9:'2018-03-04',10:'2018-03-11',11:'2018-03-18',12:'2018-03-25',13:'2018-04-01',14:'2018-04-08',15:'2018-04-15',16:'2018-04-22',17:'2018-04-29',18:'2018-05-06',19:'2018-05-13',20:'2018-05-20',21:'2018-05-27',22:'2018-06-03',23:'2018-06-10',24:'2018-06-17',25:'2018-06-24',26:'2018-07-01',27:'2018-07-08',28:'2018-07-15',29:'2018-07-22',30:'2018-07-29',31:'2018-08-05',32:'2018-08-12',33:'2018-08-19',34:'2018-08-26',35:'2018-09-02',36:'2018-09-09',37:'2018-09-16',38:'2018-09-23',39:'2018-09-30',40:'2018-10-07',41:'2018-10-14',42:'2018-10-21',43:'2018-10-28',44:'2018-11-04',45:'2018-11-11',46:'2018-11-18',47:'2018-11-25',48:'2018-12-02',49:'2018-12-09',50:'2018-12-16',51:'2018-12-23',52:'2018-12-30',53:'2019-01-06',54:'2019-01-13',55:'2019-01-20',56:'2019-01-27',57:'2019-02-03',58:'2019-02-10',59:'2019-02-17',60:'2019-02-24',61:'2019-03-03',62:'2019-03-10',63:'2019-03-17',64:'2019-03-24',65:'2019-03-31',66:'2019-04-07',67:'2019-04-14',68:'2019-04-21',69:'2019-04-28',70:'2019-05-05',71:'2019-05-12',72:'2019-05-19',73:'2019-05-26',74:'2019-06-02',75:'2019-06-09',76:'2019-06-16',77:'2019-06-23',78:'2019-06-30',79:'2019-07-07',80:'2019-07-14',81:'2019-07-21',82:'2019-07-28',83:'2019-08-04',84:'2019-08-11',85:'2019-08-18',86:'2019-08-25',87:'2019-09-01',88:'2019-09-08',89:'2019-09-15',90:'2019-09-22',91:'2019-09-29',92:'2019-10-06',93:'2019-10-13',94:'2019-10-20',95:'2019-10-27',96:'2019-11-03',97:'2019-11-10',98:'2019-11-17',99:'2019-11-24',100:'2019-12-01',101:'2019-12-08',102:'2019-12-15',103:'2019-12-22',104:'2019-12-29',105:'2020-01-05',106:'2020-01-12',107:'2020-01-19',108:'2020-01-26',109:'2020-02-02',110:'2020-02-09',111:'2020-02-16',112:'2020-02-23',113:'2020-03-01',114:'2020-03-08',115:'2020-03-15',116:'2020-03-22',117:'2020-03-29',118:'2020-04-05',119:'2020-04-12',120:'2020-04-19',121:'2020-04-26',122:'2020-05-03',123:'2020-05-10',124:'2020-05-17',125:'2020-05-24',126:'2020-05-31',127:'2020-06-07',128:'2020-06-14',129:'2020-06-21',130:'2020-06-28',131:'2020-07-05',132:'2020-07-12',133:'2020-07-19',134:'2020-07-26',135:'2020-08-02',136:'2020-08-09',137:'2020-08-16',138:'2020-08-23',139:'2020-08-30',140:'2020-09-06',141:'2020-09-13',142:'2020-09-20',143:'2020-09-27',144:'2020-10-04',145:'2020-10-11',146:'2020-10-18',147:'2020-10-25',148:'2020-11-01',149:'2020-11-08',150:'2020-11-15',151:'2020-11-22',152:'2020-11-29',153:'2020-12-06',154:'2020-12-13',155:'2020-12-20',156:'2020-12-27',157:'2021-01-03',158:'2021-01-10',159:'2021-01-17',160:'2021-01-24',161:'2021-01-31',162:'2021-02-07',163:'2021-02-14',164:'2021-02-21',165:'2021-02-28',166:'2021-03-07',167:'2021-03-14',168:'2021-03-21',169:'2021-03-28',170:'2021-04-04',171:'2021-04-11',172:'2021-04-18',173:'2021-04-25',174:'2021-05-02',175:'2021-05-09',176:'2021-05-16',177:'2021-05-23',178:'2021-05-30',179:'2021-06-06',180:'2021-06-13',181:'2021-06-20',182:'2021-06-27',183:'2021-07-04',184:'2021-07-11',185:'2021-07-18',186:'2021-07-25',187:'2021-08-01',188:'2021-08-08',189:'2021-08-15',190:'2021-08-22',191:'2021-08-29',192:'2021-09-05',193:'2021-09-12',194:'2021-09-19',195:'2021-09-26',196:'2021-10-03',197:'2021-10-10',198:'2021-10-17',199:'2021-10-24',200:'2021-10-31',201:'2021-11-07',202:'2021-11-14',203:'2021-11-21',204:'2021-11-28',205:'2021-12-05',206:'2021-12-12',207:'2021-12-19',208:'2021-12-26'},'value':{0:8057,1:8543,2:7500,3:8090,4:9858,5:9869,6:8414,7:7599,8:8805,9:8367,10:7650,11:7360,12:8205,13:7547,14:6710,15:5767,16:6902,17:7093,18:6013,19:6337,20:5172,21:6260,22:6120,23:6551,24:6363,25:6785,26:6675,27:6586,28:5872,29:5460,30:6195,31:5756,32:5856,33:5477,34:6472,35:6183,36:5732,37:5356,38:6440,39:6531,40:6473,41:6177,42:6343,43:7956,44:6439,45:6745,46:6628,47:8026,48:6859,49:6915,50:9287,51:7145,52:6082,53:5345,54:5546,55:5574,56:7117,57:6770,58:6660,59:6140,60:7521,61:6136,62:5511,63:5304,64:6010,65:5868,66:5615,67:5232,68:5202,69:5820,70:5514,71:4909,72:4558,73:5287,74:5342,75:4692,76:4599,77:4877,78:5545,79:5871,80:5430,81:4465,82:5759,83:5225,84:4832,85:4486,86:4818,87:5314,88:5070,89:4873,90:6263,91:7893,92:6766,93:7463,94:8210,95:8325,96:6934,97:6408,98:5899,99:7113,100:6919,101:7253,102:8818,103:7302,104:7541,105:6010,106:6122,107:5899,108:6794,109:6915,110:7089,111:7788,112:6756,113:6683,114:5414,115:2699,116:2075,117:2972,118:3004,119:3290,120:3620,121:3467,122:4297,123:4053,124:3803,125:4732,126:4120,127:4465,128:5465,129:4656,130:5495,131:5204,132:5005,133:4581,134:4761,135:4708,136:4375,137:4603,138:5127,139:5383,140:5162,141:5345,142:5358,143:5867,144:6885,145:6122,146:5395,147:5954,148:5318,149:5175,150:5152,151:6669,152:6654,153:5248,154:5436,155:3814,156:5276,157:4573,158:4096,159:4079,160:4574,161:4583,162:4201,163:4592,164:4880,165:5087,166:5116,167:5461,168:5337,169:5321,170:4897,171:4876,172:4756,173:4283,174:6367,175:5509,176:5856,177:5429,178:5338,179:4899,180:4094,181:4936,182:5940,183:5737,184:5471,185:5131,186:5799,187:5813,188:5209,189:5141,190:5408,191:5957,192:5442,193:5692,194:5408,195:6587,196:6365,197:6173,198:5907,199:8504,200:6358,201:7015,202:6714,203:6263,204:6826,205:6296,206:6818,207:6688,208:6895}})

Dataset2 (True):

df2 = pd.DataFrame.from_dict({'date':{0:'2017-12-31',1:'2018-01-07',2:'2018-01-14',3:'2018-01-21',4:'2018-01-28',5:'2018-02-04',6:'2018-02-11',7:'2018-02-18',8:'2018-02-25',9:'2018-03-04',10:'2018-03-11',11:'2018-03-18',12:'2018-03-25',13:'2018-04-01',14:'2018-04-08',15:'2018-04-15',16:'2018-04-22',17:'2018-04-29',18:'2018-05-06',19:'2018-05-13',20:'2018-05-20',21:'2018-05-27',22:'2018-06-03',23:'2018-06-10',24:'2018-06-17',25:'2018-06-24',26:'2018-07-01',27:'2018-07-08',28:'2018-07-15',29:'2018-07-22',30:'2018-07-29',31:'2018-08-05',32:'2018-08-12',33:'2018-08-19',34:'2018-08-26',35:'2018-09-02',36:'2018-09-09',37:'2018-09-16',38:'2018-09-23',39:'2018-09-30',40:'2018-10-07',41:'2018-10-14',42:'2018-10-21',43:'2018-10-28',44:'2018-11-04',45:'2018-11-11',46:'2018-11-18',47:'2018-11-25',48:'2018-12-02',49:'2018-12-09',50:'2018-12-16',51:'2018-12-23',52:'2018-12-30',53:'2019-01-06',54:'2019-01-13',55:'2019-01-20',56:'2019-01-27',57:'2019-02-03',58:'2019-02-10',59:'2019-02-17',60:'2019-02-24',61:'2019-03-03',62:'2019-03-10',63:'2019-03-17',64:'2019-03-24',65:'2019-03-31',66:'2019-04-07',67:'2019-04-14',68:'2019-04-21',69:'2019-04-28',70:'2019-05-05',71:'2019-05-12',72:'2019-05-19',73:'2019-05-26',74:'2019-06-02',75:'2019-06-09',76:'2019-06-16',77:'2019-06-23',78:'2019-06-30',79:'2019-07-07',80:'2019-07-14',81:'2019-07-21',82:'2019-07-28',83:'2019-08-04',84:'2019-08-11',85:'2019-08-18',86:'2019-08-25',87:'2019-09-01',88:'2019-09-08',89:'2019-09-15',90:'2019-09-22',91:'2019-09-29',92:'2019-10-06',93:'2019-10-13',94:'2019-10-20',95:'2019-10-27',96:'2019-11-03',97:'2019-11-10',98:'2019-11-17',99:'2019-11-24',100:'2019-12-01',101:'2019-12-08',102:'2019-12-15',103:'2019-12-22',104:'2019-12-29',105:'2020-01-05',106:'2020-01-12',107:'2020-01-19',108:'2020-01-26',109:'2020-02-02',110:'2020-02-09',111:'2020-02-16',112:'2020-02-23',113:'2020-03-01',114:'2020-03-08',115:'2020-03-15',116:'2020-03-22',117:'2020-03-29',118:'2020-04-05',119:'2020-04-12',120:'2020-04-19',121:'2020-04-26',122:'2020-05-03',123:'2020-05-10',124:'2020-05-17',125:'2020-05-24',126:'2020-05-31',127:'2020-06-07',128:'2020-06-14',129:'2020-06-21',130:'2020-06-28',131:'2020-07-05',132:'2020-07-12',133:'2020-07-19',134:'2020-07-26',135:'2020-08-02',136:'2020-08-09',137:'2020-08-16',138:'2020-08-23',139:'2020-08-30',140:'2020-09-06',141:'2020-09-13',142:'2020-09-20',143:'2020-09-27',144:'2020-10-04',145:'2020-10-11',146:'2020-10-18',147:'2020-10-25',148:'2020-11-01',149:'2020-11-08',150:'2020-11-15',151:'2020-11-22',152:'2020-11-29',153:'2020-12-06',154:'2020-12-13',155:'2020-12-20',156:'2020-12-27',157:'2021-01-03',158:'2021-01-10',159:'2021-01-17',160:'2021-01-24',161:'2021-01-31',162:'2021-02-07',163:'2021-02-14',164:'2021-02-21',165:'2021-02-28',166:'2021-03-07',167:'2021-03-14',168:'2021-03-21',169:'2021-03-28',170:'2021-04-04',171:'2021-04-11',172:'2021-04-18',173:'2021-04-25',174:'2021-05-02',175:'2021-05-09',176:'2021-05-16',177:'2021-05-23',178:'2021-05-30',179:'2021-06-06',180:'2021-06-13',181:'2021-06-20',182:'2021-06-27',183:'2021-07-04',184:'2021-07-11',185:'2021-07-18',186:'2021-07-25',187:'2021-08-01',188:'2021-08-08',189:'2021-08-15',190:'2021-08-22',191:'2021-08-29',192:'2021-09-05',193:'2021-09-12',194:'2021-09-19',195:'2021-09-26',196:'2021-10-03',197:'2021-10-10',198:'2021-10-17',199:'2021-10-24',200:'2021-10-31',201:'2021-11-07',202:'2021-11-14',203:'2021-11-21',204:'2021-11-28',205:'2021-12-05',206:'2021-12-12',207:'2021-12-19',208:'2021-12-26'},'value':{0:5772,1:5902,2:5820,3:5802,4:5860,5:6005,6:6052,7:6244,8:6600,9:6253,10:6099,11:6167,12:6011,13:5844,14:5956,15:5813,16:5861,17:5870,18:5877,19:5834,20:5721,21:5885,22:6013,23:5933,24:5902,25:5994,26:6157,27:6048,28:6020,29:6041,30:6103,31:5994,32:5923,33:5881,34:5934,35:5871,36:5856,37:5818,38:6010,39:5884,40:5815,41:5816,42:5880,43:5897,44:5869,45:5851,46:6007,47:5973,48:5825,49:5913,50:6136,51:5647,52:5700,53:5731,54:5681,55:5788,56:5838,57:5740,58:5665,59:5645,60:5876,61:5799,62:5749,63:5694,64:5741,65:5724,66:5848,67:5875,68:5800,69:5851,70:5941,71:5816,72:5683,73:5731,74:5839,75:5705,76:5680,77:5775,78:5866,79:5810,80:5680,81:5606,82:5709,83:5677,84:5661,85:5616,86:5633,87:5710,88:5638,89:5626,90:5612,91:5654,92:5696,93:5670,94:5625,95:5682,96:5698,97:5774,98:5734,99:5947,100:6050,101:5925,102:5956,103:5571,104:5652,105:5656,106:5642,107:5649,108:5881,109:5814,110:5774,111:5759,112:5928,113:5866,114:5973,115:5968,116:5869,117:6165,118:5875,119:5749,120:5779,121:5747,122:5876,123:5892,124:5779,125:5812,126:5742,127:5625,128:5578,129:5576,130:5589,131:5545,132:5545,133:5554,134:5553,135:5551,136:5480,137:5509,138:5610,139:5611,140:5612,141:5710,142:5839,143:5795,144:6008,145:5918,146:5840,147:5844,148:5905,149:5844,150:5818,151:5853,152:6161,153:6385,154:6303,155:5681,156:5729,157:5898,158:5873,159:5856,160:5899,161:5910,162:5926,163:6254,164:6107,165:6303,166:6229,167:6172,168:6122,169:6086,170:6394,171:6324,172:6182,173:6118,174:6351,175:6015,176:6027,177:5896,178:6061,179:6015,180:5946,181:5897,182:5913,183:5829,184:5770,185:5776,186:5865,187:5863,188:5818,189:5817,190:5811,191:5839,192:5811,193:5793,194:5792,195:6198,196:6344,197:5433,198:6192,199:6293,200:6034,201:6281,202:6138,203:6147,204:6359,205:6381,206:6485,207:6380,208:6006}})

Greykite documentation states that the seasonality "auto" option is meant to let the template decide, based on input data frequency and the amount of training data, whether to model that seasonality with default Fourier order: https://linkedin.github.io/greykite/docs/0.1.0/html/pages/model_components/0300_seasonality.html?highlight=seasonality

However, with monthly data, this option always defaults to False, both for QUARTERLY_SEASONALITY and YEARLY_SEASONALITY, even when the amount of training data (num_training_days) is greater than the minimum required (default_min_days). Why ? Read below.

These are the Silverkite default settings for minimum training data requirements, as defined in \greykite\algo\forecast\silverkite\constants\silverkite_seasonality.py

SilverkiteSeasonality(name='ct1', period=1.0, order=15, seas_names='yearly', default_min_days=548)
SilverkiteSeasonality(name='toq', period=1.0, order=5, seas_names='quarterly', default_min_days=180)

num_training_days is calculated in \greykite\common\time_properties_forecast.py, whereas the actual test is in \greykite\algo\forecast\silverkite\forecast_simple_silverkite.py(here, num_days is num_training_days calculated above):

num_days >= seas.value.default_min_days
                    and seas.name in freq_auto_seas_names

The result of the test is always False for monthly data, because freq_auto_seas_names is an empty dictionary, hence the condition seas.name in freq_auto_seas_names is never met ; the reason can be clearly seen in \greykite\algo\forecast\silverkite\constants\silverkite_time_frequency.py, where, e.g., for weekly data freq_auto_seas_names is the following dictionary:

auto_fourier_seas={SeasonalityEnum.MONTHLY_SEASONALITY.name,
                           SeasonalityEnum.QUARTERLY_SEASONALITY.name,
                           SeasonalityEnum.YEARLY_SEASONALITY.name})

whereas for monthly, quarterly and yearly data freq_auto_seas_names = {}, e.g. for monthly data:

auto_fourier_seas={
            # QUARTERLY_SEASONALITY and YEARLY_SEASONALITY are excluded from defaults
            # It's better to use `C(month)` as a categorical feature indicating the month
        })

Therefore, based on input data frequency in the first line of this issue really means: if the data frequency is one of MINUTE, HOUR, DAY, WEEK, excluding MONTH, QUARTER, YEAR, MULTIYEAR.

The "better" option in \greykite\algo\forecast\silverkite\constants\silverkite_time_frequency.py when using monthly data is thus to add an extra C(month) column as a categorical feature indicating the month.

Question: Why is this a "better" option than the following definition ?

auto_fourier_seas={SeasonalityEnum.QUARTERLY_SEASONALITY.name,
                           SeasonalityEnum.YEARLY_SEASONALITY.name})

I see the following alternatives when dealing with monthly data:

1. add an extra C(month) column as a categorical feature indicating the month; this has the disadvantage that the extra column should only be added when both QUARTERLY_SEASONALITY and YEARLY_SEASONALITY options are set to "auto" and not to "True" or "False" (quarterly and/or yearly seasonality terms are added automatically by Greykite when the respective option is set to "True", according to the valid_seas dictionary defined in _\greykite\common\enums.py; while the term in question is not added when "False")
2. Add QUARTERLY_SEASONALITY and YEARLY_SEASONALITY terms (currently excluded from defaults) to the empty auto_fouries_seas dictionary; but Greykite developers seem to prefer option 1.
3. Forget about the user setting the seasonality options ("auto", "True", "False") manually - this is applicable to all input data frequencies, not just monthly:

• [ ] Let the user configure the Fourier order and the minimum number of cycles for each seasonality
• [ ] Set the corresponding seasonality option to either "True" or "False" automatically, according to principles learned from the current logic, i.e., input data frequency, valid_seas and num_training_points >= default_min_points

One may argue that num_training_points varies between training sets when using CV splits; however, the following example shows that both num_training_points and num_training_days are invariant between splits, even with cv_expanding_window =True:

[CV 1/3] ... valid_seas={'YEARLY_SEASONALITY', 'QUARTERLY_SEASONALITY'})>, 'num_training_points': 26, 'num_training_days': 789.0, 'days_per_observation': 28.0, ...
[CV 2/3] ... valid_seas={'YEARLY_SEASONALITY', 'QUARTERLY_SEASONALITY'})>, 'num_training_points': 26, 'num_training_days': 789.0, 'days_per_observation': 28.0, ...
[CV 3/3] ... valid_seas={'YEARLY_SEASONALITY', 'QUARTERLY_SEASONALITY'})>, 'num_training_points': 26, 'num_training_days': 789.0, 'days_per_observation': 28.0, ...

This means that the test num_training_points >= default_min_points can be applied only once directly from train_end_date before entering the CV loop (the current Fitting 3 folds for each of 1 candidates, totalling 3 fits section apparently tests the seasonality terms at each split, but the test values are invariant, as mentioned above).

Hi, I was wondering if it is possible to extract the different modeling components (e.g. trend, holidays, seasonalities) from the forecasted time series. It's possible to do this in the Prophet framework, see: https://github.com/facebook/prophet/issues/1920

The reason is that I would like to use a custom trend component calculated outside of Greykite.