Amazon Personalize Movie Recommendation Example

This example workflow builds a recommendation model using the Amazon Personalize web services. It uses the movielens dataset which contains information about users, movies (items) and interactions of users with movies, i.e., which user has watched which movie at which time and how did he or she rate that movie. The first step is to upload the data to Amazon. We filter all the interactions out that have a rating less than 4 (out of 5) because we want the model later on to recommend only movies that the user will like. With this data, a user personalization model (solution version) is built and afterwards deployed as campaign. We can then use this campaign to make personalized movie recommendations.

This workflow will run about one or two hours. It follows the example of the blog article, Amazon Personalize Real-Time Personalization and Recommendation for Everyone. See the link below to the specific article.

Note: You need to have AWS credentials and authenticate with the Amazon Authentication node in order to use the Amazon Personalize nodes.

External resources

• KNIME and AWS Machine Learning Services Integration
• Amazon Personalize Real-Time Personalization and Recommendation for Everyone
• Amazon Personalize