Prophet is an additive regression model with the following components:
where g(t) - trend,
s(t) - seasonality,
h(t) - holidays or other business impactive events,
ϵₜ - irreducible error term
This model is specialized in time-series analysis and forecasting.
Bricks → Machine Learning → Prophet
Three options are available: yearly, weekly, daily. It is possible to choose all three of them at the same time or only those most needed.
- Date column
The column contains a date range and should be in a DateTime format.
- Target variable
The column contains target values for modeling.
- Filter Columns
If you have columns in your data that need to be ignored (but not removed from the data set) and not be shown in the model, you should specify them in this parameter. To select multiple columns, click the '+' button in the brick settings.
In addition, you can ignore all columns except the ones you specified by enabling the "Remove all except selected" option.
Brick takes the dataset with time-series data.
Brick returns the result of the model. For more information on the model, results click on the Open View button.
Prophet is used to build a model for time series. For this model not only seasonality variables may be included but also others that may be considered as important.
Let’s use it to build a model for predicting sales based on previous data from the stores. Firstly, we aggregated the data by date.
Next, we use Prophet brick in which we choose all of the seasonal variables and disallowed negative predictions since sales cannot be negative.
As a result, we have a model which may need some improvement but overall it is ready to make some predictions.
We use Predict brick and data which we generated with Date Range brick to make prediction based on our Prophet model.