Amazon SageMaker Canvas announces new capabilities for time series forecasting models
Amazon SageMaker Canvas announces new capabilities to build, evaluate, and deploy time series forecasting models, providing greater flexibility and ease of use to build forecasting applications. Amazon SageMaker Canvas is a no-code workspace that empowers analysts and citizen data scientists to build, customize, and deploy machine learning (ML) models to generate accurate predictions.
To build time series forecasting models, SageMaker Canvas uses up to six built-in algorithms to create a custom ensemble of models for each item in your time series, resulting in highly accurate models. Starting today, SageMaker Canvas provides visibility into these algorithms and the flexibility to choose any combination of these algorithms to build your time series forecasting model. Once the model is built, SageMaker Canvas provides a leaderboard with a ranked list of model candidates including a recommendation for the best model based on your dataset and the problem to be solved. You can review key performance metrics for each model on the leaderboard and select a model of your choice. The selected model can then be deployed into production on an Amazon SageMaker real-time inference endpoint for use in applications outside SageMaker Canvas.
To access the algorithm selection, model leaderboard, and direct deployment to real-time endpoint capabilities for time series forecasting, log out and log back in to SageMaker Canvas. The new capabilities are now available in all AWS regions where SageMaker Canvas is supported. To learn more, refer to the SageMaker Canvas product documentation.