Amazon’s AWS today launched Amazon Forecast, a new pre-built machine learning tool that will make it easier for developers to generate predictions based on time-series data. While predictions are pretty much the most standard use case for machine learning, building them still takes some skill. Amazon, of course, has already built plenty of these models for its own needs, so now it is essentially turning them into a product.
“With just three clicks, you can give us the information and get a forecast,” AWS CEO Andy Jassy said in today’s AWS re:Invent keynote. “It’s super simple and when we benchmarked with customers in the private beta and ourselves, it’s providing up to 50 percent more accurate forecasts than what people were doing on their own before at one-tenth of the cost of traditional supply chain software.”
Amazon, in its retail business, built a number of models to handle its own data. This is essentially the same technology that Amazon uses to forecast demand on its retail site. Users provide the company with all of their supply chain data and then give the service the variables that could have impact on the forecast.
Behind the scenes, AWS looks at the data and the signal and then chooses from eight different pre-built algorithms, trains the model, tweaks it and provides the forecast.
AWS is also making it easy to integrate this service with SAP’s and Oracle’s supply chain tools, as well as Amazon’s new Timestream database service.
The service isn’t necessarily cheap, but it will surely save developers a lot of time.
