In November, Amazon Web Services (AWS) made available to a limited group of developers its machine learning service for creating individualized recommendations, called Personalize.
The tech giant said it provided essentially the same sophisticated recommendation service that the Amazon store employs, for personalized product and content recommendations, tailored search results and targeting marketing promotions.
On Monday, AWS announced the general availability rollout of Personalize. The first AWS regions to have access are northern Virginia, Ohio, Oregon, Toyko, Singapore and Ireland. Amazon points out that its recommendation service – arguably the best known, along with Netflix’s – isn’t a master algorithm, but a mix of data, algorithms and optimizations for each use case.
To utilize this personalization-as-a-service, a publisher provides an activity stream from an application, which can include such data as clicks, page views, signups or purchase history – along with info on the products to be recommended, such as products, videos, songs or articles.
Additional user info, including demographic or geographic data, can also be included. AWS said the supplied data is kept private and secure, and only used for that application’s recommendations.
The service selects the most appropriate algorithms, trains a personalized machine learning model that is designed for the data, and then hosts and manages the model as it provides the recommendations via an API call. Application owners can control the service through the AWS console, and billing is only for the amount of the service used, with no minimums or upfront fees.
Amazon said that example use cases could include recommendations to individual users of a video streaming website, based on past viewing habits and demographics. The company cites Daniel Muller, Head of Cloud Infrastructure at Over-the-Top TV service Spuul, who said that his company took only three days to begin using the service, even though it had no prior experience in AI or machine learning.
The recommendation service can also be used to help personalize search, by ranking search results on the basis of past interactions with that user. Amazon gives the example of an ecommerce retailer who can personalize search results based on a shopper’s recent views, purchase history and preferences.
An executive at wedding site Zola.com is quoted as saying that, previously, the site utilized rule-based ranking, popularity or a similarity model, but now it can employ machine learning to find patterns.
Personalize is also designed to guide the most relevant marketing promotions, such as sending the most appropriate mobile app notification based on location, buying habits and the most effective deals for a particular type of customer.
The post Amazon’s famed recommendation service Personalize now available to every application appeared first on ClickZ.Reblogged 2 months ago from www.clickz.com