Greykite models and components¶
See Choose a Model for model options and their comparison.
In Configure a Forecast,
follow Choose a Model Template
to set config.model_template
to use the proper model.
Silverkite¶
Silverkite is a forecasting algorithm developed by LinkedIn.
It works by generating basis functions for growth, seasonality, holidays, etc. These features, along with any regressors you provide, are used to fit the timeseries.
The features can be combined with interaction terms in a flexible and powerful way. A single model is used for fitting.
This approach has the following advantages.
Flexible. Supports different kinds of growth, interactions, and fitting algorithms.
Intepretable. The default fitting algorithms are additive so you can identify the contribution of each component.
Fast. Runs much faster than Bayesian alternatives.
model_components¶
config.model_components_param
is an instance of
ModelComponentsParam
.
The attributes represent different categories of tuning parameters. See the rest of this section for how to configure each component.
from greykite.framework.templates.autogen.forecast_config import ModelComponentsParam
model_components = ModelComponentsParam(
growth=growth,
seasonality=seasonality,
events=events,
changepoints=changepoints,
regressors=regressors,
uncertainty=uncertainty,
custom=custom,
hyperparameter_override=hyperparameter_override
)