AWS brings no-code to Amazon SageMaker machine learning

Amazon SageMaker Canvas provides a visual interface for preparing data and training models, no programming or machine learning expertise required.

Amazon Web Services (AWS) has announced limited general availability of Amazon SageMaker Canvas, a visual, no-code tool for creating machine learning models aimed at business analysts.

Built as a new capability for the Amazon SageMaker machine learning service, SageMaker Canvas provides a visual interface that accesses data from disparate sources and prepares the data for training machine learning (ML) models. A point-and-click interface enables generation of accurate ML predictions, without requiring ML experience or writing any code. SageMaker Canvas is integrated with with Amazon SageMaker Studio.

Amazon SageMaker uses AutoML technology to train models based on a given dataset. SageMaker cleans and combines the data, creates hundreds of models, and selects the best one. Individual or batch predictions are generated. 

Use cases can be addressed such as fraud detection, churn reduction, and inventory optimisation. Multiple machine learning problem types are supported including binary and multi-class classifications, numerical regression, and time series forecasting.

Data can be accessed from cloud-based and on-premises data sources. SageMaker Canvas corrects data errors and analyses data readiness for ML. But as of November 30 SageMaker Canvas was available only in Oregon, Ohio, and Northern Virginia in the US, and in Frankfurt, Germany, and Ireland.

AWS has also unveiled a preview of Amazon SageMaker Studio Lab, a free service to experiment with machine learning. Amazon SageMaker Studio Lab is based on the open source JupyterLab notebook interface and offers free access to AWS compute resources.