Data Input and Output


Applies to: Machine Learning Studio (classic)

This content pertains only to Studio (classic). Similar drag and drop modules have been added to Azure Machine Learning designer (preview). Learn more in this article comparing the two versions.

This article lists the modules that you can use for importing and exporting data and models in Azure Machine Learning Studio (classic).

In addition to using modules, you can directly upload and download datasets from local files on your computer or network. For more information, see Upload existing data into an Azure Machine Learning experiment.

Here are some of the sources you can use to import and export data and models in Machine Learning Studio (classic):

  • Get data from sources in the cloud, such as Azure SQL Database, Azure SQL Data Warehouse, Azure Storage, and Azure Cosmos DB. You can also import data that's provided as a public web URL, get data from Hadoop by using a Hive query, or query an on-premises SQL server.
  • Load a collection of images from Azure Blob storage to use in image classification tasks.
  • Extract the data from zipped files that you uploaded to Machine Learning. You can use the datasets in experiments.
  • Create small datasets by typing in the Machine Learning Studio (classic) UI. This can be handy for creating small test datasets.
  • Save your results or intermediate data to Azure Table storage, Blob storage, a SQL database, or a Hive query.
  • Get a trained model from a URL or Blob storage, and then use it in an experiment.


The modules in this group only move data to or from Machine Learning Studio (classic). You can't use the modules to filter, cast, or transform the data during the import or export process.

For more information about how to transform and filter your data in Machine Learning Studio (classic), see Data Transformation.


The following articles introduce common data scenarios in machine learning:

Get started

Learn how to manage data for machine learning in the cloud. The information in this article is based on CRISP-DM, an industry standard. The article provides end-to-end walkthroughs that demonstrate the integration of machine learning with cloud data solutions such as Azure HDInsight and SQL Database.

This article describes how to get your data into Azure, and then create an experiment.

Learn how to use ETL tools to integrate your data warehouse with Machine Learning.

Advanced data science

Learn how to install the Machine Learning Python client library, and then use it to access metadata and work with datasets.

Sample experiments

List of modules

The Data Input and Output category includes the following modules:

  • Enter Data Manually: Lets you create small datasets by typing values.
  • Export Data: Writes a dataset to web URLs or to various forms of cloud-based storage in Azure, such as tables, blobs, or a SQL database.
  • Import Data: Loads data from external sources on the web and from various forms of cloud-based storage in Azure, such as Table storage, Blob storage, SQL Database, SQL Data Warehouse, Azure Cosmos DB, or a Hive query. You can also import data from an on-premises SQL Server database.
  • Load Trained Model: Gets a trained model from a URL or Blob storage to use in a scoring experiment.
  • Unpack Zipped Datasets: Decompresses a dataset that's been stored in zipped format, and then adds the dataset to your workspace.

See also