Amazon Personalize
Overview
Amazon Personalize draws on features that Amazon incorporates into their own retail website. This includes personalization experiences, including specific product recommendations, personalized product re-ranking, and customized direct marketing. The Amazon Personalize AI service provides personalisation with AWS doing all the heavy lifting of providing the Machine Learning infrastructure to train and deploy the model. If more control over the algorithm was need a Sagemaker built in algorithm could be used. The two algorithms most commonly used for recomendation problems are:
- AWS docs: https://aws.amazon.com/personalize/
- AWS FAQs: https://aws.amazon.com/personalize/faqs/
Video: Improve Customer Engagement and Conversion with Amazon Personalize
Key features
- Real-time or batch recommendations
- New user and new item recommendations
- Contextual recommendations
- Similar item recommendations
Personalize takes business data and uses it to train Machine Learning models. There are three types of data:
User interactions are records of events caused by users interacting with the business, for example by using the business website. This could be what they viewed or purtchased etc. Item metadata could be a product catalogue with data about each product. User metadata could be age, gender, location etc.
- User interactions
- Item metadata
- User metadata
Use cases
- Personalized recommendations
- Similar items
- Personalized reranking i.e. rerank a list of items for a user
- Personalized promotions/notifications
The example above is from the Amazon website. The PHP in easy steps book was previously viewed, so it has been re-displayed with other books from the in easy steps series. The previous viewing was the user interaction and the books displayed is an example of displaying similar items.
Video: Introduction to Amazon Personalize
Credits
- Photo by freestocks on Unsplash
