Python Forecaster (Python Scripting)

Description

The Python Forecaster operator allows you to create custom forecast models using Python scripts within Altair AI Studio. Depending on the operator's configuration, you can edit the Python script to define the model's behavior according to your specific needs.

If the operator is editable, the parameter panel includes buttons at the top of the parameter list that allow you to:

- **Edit the Operator Declaration**: Customize capabilities, parameter types, and configurable input and output ports.

- **Edit the Python Script**: Modify the script that trains and applies your forecast model.

- **Save as Template**: Save your configuration for future use.

You can define user-configurable parameters to make your operator more flexible. To create a new parameter, extend the editable JSON parameters array by adding a new JSON object like the following: { "name": "my_new_parameter", "description": "This is a description what the parameter can be used for.", "type": "integer", "optional": false, "value": 100 }

Similarly, you can define user-configurable input ports. To create a new input port, extend the editable JSON inputs array by adding a new JSON object: { "name": "my_new_input_port", "type": "table" } The new input can be easily reached by adding an extra attribute to the rm_train method. For example: rm_train(index, series, new_input_data, parameters)

You can also define user-configurable output ports. To create a new output port, extend the editable JSON outputs array by adding a new JSON object: { "name": "my_new_output_port", "type": "table" }

The Python script requires two functions: rm_train and rm_apply.

The rm_train function is responsible for training the model. It should return an object that contains the trained model and any other data needed by the rm_apply function. If you have defined user-configurable output ports (e.g., to output performance metrics), you should return them along with the model object. For instance: return model_object, pd.DataFrame({"performance": [performance_metric]})

If you have multiple output ports, return additional DataFrames for each: return model_object, dataframe_output1, dataframe_output2

Note that the parameters variable in the rm_train function includes the names of the time series attribute (series_name) and the index (index_name) by default.

The rm_apply function is called by the Apply Forecast operator to perform forecasting. The model argument contains the Python object returned by the rm_train function. This function should return the forecasted values as a pandas DataFrame.

Please note that the model generated by this operator wraps the Python model returned by your Python script. Both training and applying the model require a compatible Python environment.

The Apply Forecast operator uses the trained forecast model to generate forecasts for the time series it was trained on.