Segmented Modeling Projects

Many time series multiseries projects introduce complex forecasting use cases that require using different models for subsets of series (i.e., sales of groceries and clothing can be very different). Within the segmented modeling framework, DataRobot runs multiple time series projects (one per segment / group of series), selects the best models for each segment, and then combines those models to make predictions.


A segment is a group of series in a multiseries project. For example, given store and country columns in dataset, you can use the former as the series identifier and the latter as the segment identifier. For the best results, group series with similar patterns into segments (instead of random selection).

Segmentation Task

A segmentation task is an entity that defines how input dataset is partitioned. Currently only user-defined segmentation is supported. That is, the dataset must have a separate column that is used to identify segment (and the user must select it). All records within a series must have the same segment identifier.

Combined Model

A combined model in a segmented modeling project can be thought of as a meta-model made of references to the best model within each segment. While being quite different from a standard DataRobot model in its creation, its use is very much the same after the model is complete (for example, deploying or making predictions).

The following examples illustrate how to set up, run, and manage a segmented modeling project using the Python public API client. For details please refer to Segmented Modeling API Reference.

Starting a Segmentation Project with a User Defined Segment ID

Time series modeling must be enabled for your account to run segmented modeling projects.

Use the standard method to create a DataRobot project:

from datarobot import DatetimePartitioningSpecification
from datarobot import enums
from datarobot import Project
from datarobot import SegmentationTask

project_name = "Segmentation Demo with Segmentation ID"
project_dataset = "multiseries_segmentation.csv"
project = Project.create(project_dataset, project_name=project_name)

datetime_partition_column = "timestamp"
multiseries_id_column = "series_id"
user_defined_segment_id_column = "物类segment_id"
target = "target"

Create a simple datetime specification for a time series project:

spec = DatetimePartitioningSpecification(

Create a segmentation task for the project:

segmentation_task_results = SegmentationTask.create(,
segmentation_task = segmentation_task_results["completedJobs"][0]

Start a segmented project by passing the segmentation_task_id argument:


Working with Combined Models

Retrieve Combined Models:

from datarobot import Project, CombinedModel
project_id = "60ff165dde5f3ceacda0f2d6"

# Get an existing segmentation project
project = Project.get(segmented_project_id)

# Retrieve list of all combined models in the project
combined_models = project.get_combined_models()

# Or just an active (current) combined model
current_combined_model = project.get_active_combined_model()

Get information about segments in the Combined Model:

segments_info = current_combined_model.get_segments_info()

# Alternatively this information can be retrieved as a Pandas DataFrame
segments_df = current_combined_model.get_segments_as_dataframe()

# Or even in CSV format

Ensure Autopilot has completed for all segments:

segments_info = current_combined_model.get_segments_info()
assert all(segment.autopilot_done for segment in segments_info)

Optionally, view a list of all models associated with individual segments:

segments_and_child_models = project.get_segments_models(

Set a new champion for a segment in the Combined Model, specifying the project_id of the segmented project and the model_id from that project:

segment_project_id = "60ff165dde5f3ceacdaabcde"
new_champion_id = "60ff165dde5f3ceacdaa12f7"

CombinedModel.set_segment_champion(project_id=segment_project_id, model_id=new_champion_id)

If active Combined Model has already been deployed - changing champions is not allowed. In this case, create a copy of Combined Model, make it active, and set champion for it (deployed model remains unchanged):

new_combined_model = CombinedModel.set_segment_champion(project_id=segment_project_id, model_id=new_champion_id, clone=True)

Run predictions on the Combined Model:

prediction_dataset = "multiseries_predictions.csv"

# Upload dataset
dataset = project.upload_dataset(

# Request predictions
predictions_job = current_combined_model.request_predictions(,
predictions = predictions.get_result()