Components breakdown for forecast

[vandonova]

Hi,

I'm looking to break down my forecast into its components. I saw your function plot_components however, it only uses the train data to create the components.

In my example, my training data never has values on Saturdays. In that case,

params = {'forecast_horizon': 1,
 'fit_algorithm_dict': {'fit_algorithm': 'elastic_net',
  'fit_algorithm_params': {'l1_ratio': [0.1, 0.5, 0.7, 0.9, 0.95, 0.99, 1]}},
 'changepoints_dict': {'method': 'auto',
  'regularization_strength': None,
  'actual_changepoint_min_distance': '7D',
  'potential_changepoint_distance': '3D',
  'yearly_seasonality_order': 6},
}
model = SimpleSilverkiteEstimator(**params)
model.fit(data)
predictions = model.predict(data)
model.plot_components()

return a weekly component plot that just connects Friday and Sunday with a straight line. (Note that hovering on 5, Saturday, does not show any data).

When predicting, I want to include Saturdays in my forecast and predict their value based on the fitted coefficients for 'sin1_tow_weekly', 'cos1_tow_weekly', 'sin2_tow_weekly', 'cos2_tow_weekly', 'sin3_tow_weekly', 'cos3_tow_weekly'.

Is there an easy way to retrieve the components for predictions? Alternatively, how would I access the design matrix x_mat for the forecast instead of the fitted data?

Thank you in advance!
Vicky

[sayanpatra]

Hi @vandonova,

You can retrieve the components for training as follows:

fig = result.model[-1].plot_components()  # Generates and stores the components
result.model[-1].silverkite_diagnostics.components

Currently there is no easy way to return the components for the forecasts, but it can be computed as follows.

Assumption: result is the output from the pipeline

import patsy

estimator = result.forecast.estimator
trained_model= estimator.model_dict
fut_df = result.forecast.df[["ts", "actual"]].rename(columns={"actual":"y"})

# Build Silverkite features
features_df_fut = estimator.silverkite._SilverkiteForecast__build_silverkite_features(
    df=fut_df,
    time_col=trained_model["time_col"],
    origin_for_time_vars=trained_model["origin_for_time_vars"],
    daily_event_df_dict=trained_model["daily_event_df_dict"],
    changepoint_values=trained_model["changepoint_values"],
    continuous_time_col=trained_model["continuous_time_col"],
    growth_func=trained_model["growth_func"],
    fs_func=trained_model["fs_func"],
    seasonality_changepoint_result=trained_model["seasonality_changepoint_result"],
    changepoint_dates=trained_model["trend_changepoint_dates"]
)

# Add autoregression terms
if trained_model["autoreg_func"] is not None:
    past_df = trained_model["df"].copy()
    df = pd.DataFrame({value_col: [np.nan]*fut_df.shape[0]})
    df.index = fut_df.index
    autoreg_df = estimator.silverkite._SilverkiteForecast__build_autoreg_features(
        df=df,
        value_col=trained_model["value_col"],
        autoreg_func=trained_model["autoreg_func"],
        phase="predict",
        past_df=past_df[[value_col]])
    features_df_fut = pd.concat(
        [features_df_fut, autoreg_df],
        axis=1,
        sort=False)

# Compute the design matrix
(x_mat_predict,) = patsy.build_design_matrices(
    [trained_model["x_design_info"]],
    data=features_df_fut,
    return_type="dataframe")

# Compute features df ( this is `x_mat_predict` multiplied by the training coefficients)
feature_df_predict = (trained_model["ml_model"].coef_* x_mat_predict)

# Calculate components
diagnostics = estimator.silverkite_diagnostics
components_predict = diagnostics.get_silverkite_components(
                fut_df,
                diagnostics.time_col,
                diagnostics.value_col,
                feature_df_predict)

Hope this helps. Have a good weekend.

[vandonova]

Thanks @sayanpatra this is exactly what I needed.