Amazon Machine Learning is a managed service for building ML models and generating predictions, enabling the development of robust, scalable smart applications. Amazon Machine Learning enables you to use powerful machine learning technology without requiring an extensive background in machine learning algorithms and techniques.

The process of building ML models with Amazon Machine Learning consists of three operations: data analysis, model training, and evaluation. The data analysis step computes and visualizes your data’s distribution, and suggests transformations that optimize the model training process. The model training step finds and stores the predictive patterns within the transformed data. In the optional final step, the model is evaluated for accuracy.

Amazon Machine Learning combines powerful machine learning algorithms with interactive visual tools to guide you towards easily creating, evaluating, and deploying machine learning models. Its built-in data transformations ensure that input datasets can be seamlessly transformed to maximize the model's predictive quality. Once a model is built, the service's intuitive model evaluation and fine-tuning console help you understand its strengths and weaknesses, and adjust its performance to meet business objectives.

Get started with Amazon Machine Learning

Create a Free Account

Receive twelve months of access to the AWS Free Tier and enjoy AWS Basic Support features including, 24x7x365 customer service, support forums, and more.

Please note that Amazon Machine Learning is not currently available on the AWS Free Tier.


Machine learning (ML) can help you use historical data to make better business decisions. ML algorithms discover patterns in data and construct predictive models using these patterns. Then, you can use the models to make predictions on future data. For example, one possible application of ML would be to predict whether or not a customer will purchase a particular product based on past behavior, and use this prediction to send a personalized promotional email to that customer.


Amazon Machine Learning makes it easy to work with data that is already stored in the AWS cloud. You can use datasets that are already stored as CSV files in Amazon S3, or query Amazon Redshift or MySQL databases in Amazon RDS to create and use ML models.


High-quality data is critical to building accurate predictive models, but real world datasets are frequently incomplete or inconsistent. Amazon Machine Learning provides interactive charts that help you visualize and explore your input datasets to understand data content and distribution and spot missing or incorrect data attributes.

Click to Enlarge

Click to Enlarge


Amazon Machine Learning makes it easy to understand your model’s performance by calculating industry-standard quality metrics and providing visualization of the models behavior. Amazon Machine Learning can also help you fine-tune the interpretation of the predictions. For example, if your ML model is used to classify purchases as legitimate or fraudulent, Amazon Machine Learning will help you visualize prediction results, and decide how to adjust the predictions to deliver the optimal results for your smart application.

Click to Enlarge

Click to Enlarge

Click to Enlarge

Amazon Machine Learning provides APIs for modeling and management that allow you to create, review, and delete data sources, models, and evaluations. This allows you to automate the creation of new models when new data becomes available. You can also use the APIs to inspect previous models, data sources, evaluations, and batch predictions for tracking and repeatability.


Amazon Machine Learning uses scalable and robust implementations of industry-standard ML algorithms. With Amazon Machine Learning, developers can create models that predict values of binary attributes (binary classification), categorical attributes (multi-class classification), or numeric attributes (regression). For example, a binary classification model can be used to predict whether a website comment is spam (e.g., yes or no). Multi-class classification models can be used to predict where to route a customer service request (e.g., "Billing", "Technical Support", or "Order status"). Regression models can be used to predict the number of days before a customer's next interaction with an application or service.


The quality of your machine learning models depends on the quality of the input data and how the data is transformed before being input to the ML algorithm. To help you get the most from your data, Amazon Machine Learning provides implementations of common ML data transformations. Amazon Machine Learning will automatically suggest data transformations for your input data, and you can easily adjust which transformations are applied to attributes in your data at model training time.


Once you create your machine learning models, Amazon Machine Learning provides APIs to obtain predictions from them, enabling you to easily build smart applications. It can generate billions of predictions using the batch prediction API, or serve predictions at high throughput and with low latency with the real-time API. The batch prediction API retrieves a large number of data records and generates predictions all at once while the real-time prediction API generates predictions synchronously and with low-latency.


Amazon Machine Learning manages all the infrastructure and workflows needed to run and scale ML model creation and prediction generation, freeing you to focus on your application. You can create as many models as you need, and scale the volume and throughput of predictions generated by these models without worrying about provisioning hardware, distributing and scaling the computational load, managing dependencies, or monitoring and troubleshooting your ML fleet.


With Amazon Machine Learning you pay only for what you use, making it easy and cost effective to scale from generating a few predictions a day to hundreds per second. You are charged an hourly rate for the compute time used to build predictive models and a per-prediction rate for both batch and real-time predictions. You are also charged for real-time predictions based on the amount of memory required for each model.