Testing External Datasets in DataRobot

This post was originally part of the DataRobot Community. Visit now to browse discussions and ask questions about DataRobot, AI Cloud, data science, and more.

This article showcases how you can upload your own external testing datasets into DataRobot to evaluate the performance of your models.

Overview

DataRobot will handle partitioning automatically, ensuring that models are evaluated on out-of-sample data. However, analysts are still given the option to upload any number of additional test datasets and compare metric scores to ensure consistency prior to deployment.

Note: To ensure that you can access this functionality, contact your DataRobot representative for information on enabling the feature.

Uploading External Dataset

To upload the external dataset, navigate to the Make Predictions page by clicking the Predict tab for your DataRobot model (Figure 1).

Use one of the Import data from options and upload your dataset. When the dataset finishes uploading, you see the option “Run External Test” (Figure 2). Click that link.

DataRobot takes a few moments to finish calculating the accuracy metrics against this dataset for the related model.

Sorting by external test data accuracy

To see and use the newly calculated accuracy, select Menu > Show External Test Column (Figure 3).

You can now sort models by external test scores and calculate scores for more models.

Caveats

Currently, external test datasets are supported for:

• Regression
• Binary Classification
• Multiclass Classification
• OTV projects

More information

Search DataRobot documentation for Testing with external datasets.

About the author

Linda Haviland

Community Manager

Meet Linda Haviland