AutoTSModel#

class datarobotx.AutoTSModel(name=None, feature_window=None, forecast_window=None, **kwargs)[source]#

AutoTS orchestrator

Applies automatic feature engineering and a model selection process optimized for the nuances of time-series modeling (e.g. automatic engineering of lag variables, handling of varying forecast horizons).

Parameters:

name (str, optional) – Name to use for the DataRobot project that will be created. Alias for the DR ‘project_name’ configuration parameter.
feature_window (tuple of int, optional) –
Window of time relative to the forecast point over which to automatically derive features. Expected format is a tuple: (start, end) e.g. (-20, 0). Larger windows require more data for predictions. Window unit of time is automatically detected but can be explicitly specified separately.

Alias for the DR ‘feature_derivation_window_start’ and ‘feature_derivation_window_end’ configuration parameters.
forecast_window (tuple of int, optional) –
Window of time relative to the forecast point for which predictions are of interest. Expected format is a tuple: (start, end) e.g. (1, 5). Window unit of time is automatically detected but can be explicitly specified separately.

Alias for the DR ‘forecast_window_start’ and ‘forecast_window_end’ configuration parameters.
**kwargs – Additional DataRobot configuration parameters for project creation and autopilot execution. See the DRConfig docs for usage examples.

See also

DRConfig: Configuration object for DataRobot project and autopilot settings, also includes detailed examples of usage

Inherited attributes:

`dr_model`	DataRobot python client datarobot.Model object for the present champion
`dr_project`	DataRobot python client datarobot.Project object

Methods:

fit(X, datetime_partition_column[, target, ...])

Fit time-series challenger models using DataRobot

Inherited methods:

`deploy`([wait_for_autopilot, name])	Deploy the model into ML Ops
`from_project_id`(project_id)	Class method to create from an existing project id
`from_url`(url)	Class method to initialize from a URL string
`get_params`()	Configuration parameters for the model
`predict`(X[, wait_for_autopilot])	Make batch predictions using the present champion
`predict_proba`(X[, wait_for_autopilot])	Calculate class probabilities using the present champion
`set_params`(**kwargs)	Set or update configuration parameters for the model
`share`(emails)	Share a project with other users.

deploy(wait_for_autopilot=False, name=None)[source]#

Deploy the model into ML Ops

Return type:

Deployment

Returns:

Deployment – Resulting ML Ops deployment
wait_for_autopilot (bool, optional, default=False) – If True, wait for autopilot to complete before deploying the model In non-notebook environments, fit() will always block until complete
name (str, optional, default=None) – Name for the deployment. If None, a name will be generated

property dr_model: datarobot.Model#

DataRobot python client datarobot.Model object for the present champion

Returns:: datarobot.Model object associated with this drx model
Return type:: datarobot.Model

property dr_project: datarobot.Project#

DataRobot python client datarobot.Project object

Returns:: datarobot.Project object associated with this drx.Model
Return type:: datarobot.Project

fit(X, datetime_partition_column, target=None, multiseries_id_columns=None, **kwargs)[source]#

Fit time-series challenger models using DataRobot

Automatically engineers temporally lagged features and establishes time-series champion models across a forecast horizon of interest.

Parameters:

X (pandas.DataFrame) – Training dataset for challenger models. If str, can be AI catalog dataset id or name (if unambiguous)
target (str, default=None) – Column name from the dataset to be used as the target variable. If None, TS anomaly detection will be executed.
datetime_partition_column (str) – Column name from the dataset containing the primary date/time feature to be used in partitioning, feature engineering, and establishing forecast horizons
multiseries_id_columns (str or list of str, optional) – Column name from the dataset containing a series identifier for each row. If specified, DataRobot will treat this as a multiseries problem. Presently only a single series id column is supported.
**kwargs – Additional optional fit-time parameters to pass to DataRobot i.e. ‘weights’

See also

DRConfig: Configuration object for DataRobot project and autopilot settings, also includes detailed examples of usage.

Return type:: AutoTSModel

classmethod from_project_id(project_id)[source]#

Class method to create from an existing project id

Initializes a new object from the provided project_id. Configuration parameters originally used to create the project and start Autopilot may not be recoverable.

Parameters:: project_id (str, optional) – DataRobot id for the project from which to initialize the object
Returns:: model – New AutopilotModel instance
Return type:: AutopilotModel

Examples

>>> my_model = AutopilotModel.from_project_id('62f14505bab13ab73593d69e')

classmethod from_url(url)[source]#

Class method to initialize from a URL string

Useful for copy and pasting between GUI and notebook environments

Parameters:: url (str) – URL of a DataRobot GUI page related to the project of interest
Returns:: model – The constructed AutopilotModel object
Return type:: AutopilotModel

get_params()[source]#

Configuration parameters for the model

Note that some parameters may be initialized or materialized server-side after creating a project or starting Autopilot. get_params() only returns the client-side parameters which will be (or were) passed to DataRobot.

Returns:: config – Configuration object containing the parameters to be used with DataRobot
Return type:: DRConfig

predict(X, wait_for_autopilot=False)[source]#

Make batch predictions using the present champion

Predictions are calculated asynchronously - returns immediately but reinitializes the returned DataFrame with data once predictions are completed.

Predictions are made within the project containing the model using modeling workers. For real-time predictions, first deploy the model.

Parameters:

X (pandas.DataFrame) – Dataset to be scored - target column can be included or omitted
wait_for_autopilot (bool, optional, default=False) – If True, wait for autopilot to complete before making predictions In non-notebook environments, fit() will always block until complete

Returns:

Resulting predictions (contained in the column ‘predictions’) Returned immediately, updated automatically when results are completed.

Return type:

FutureDataFrame

predict_proba(X, wait_for_autopilot=False)[source]#

Calculate class probabilities using the present champion

Only available for classifier and clustering models.

Parameters:

X (pandas.DataFrame) – Dataset to compute class probabilities on; target column can be included or omitted
wait_for_autopilot (bool, optional, default=False) – If True, wait for autopilot to complete before making predictions In non-notebook environments, fit() will always block until complete

Returns:

Resulting predictions; probabilities for each label are contained in the column ‘class_{label}’; returned immediately, updated automatically when results are completed.

Return type:

FutureDataFrame