AWS launches machine learning service

Amazon is challenging Microsoft and IBM with the announcement of a new cloud-based machine learning service.

The company claims Amazon Machine Learning on Amazon Web Services will allow developers without deep data science skills to take advantage of machine learning (ML).

Microsoft launched its own Azure Machine Learning Service last year, on the heels of IBM's move to separate its Watson system as a cloud-based business unit.

The AWS service, launched at its annual summit in San Francisco and available today, employs the same ML technology "used for years by Amazon’s internal data scientist community". The company suggests use cases could range from fraud detection to modelling and prediction of customer behaviour.

A blog post by AWS' chief evangelist Jeff Barr includes a step-by-step demonstration of how developers with existing unexploited datasets would use the new service and offers suggestions on preparing data for the training process.

Pricing is on a pay-as-you-go basis with data analysis, model training, and model evaluation priced at US$0.42 (A$0.55) per compute hour.

Batch predictions will cost US$0.10 (A$0.13) for every 1000 predictions, rounded up to the next 1000, with real-time predictions incurring an additional hourly reserved capacity charge.

Charges for data stored in S3, Amazon RDS, and Amazon Redshift are billed separately.