Which solution will meet these requirements with the LEAST development effort?
Train a model by using a user-based collaborative filtering algorithm on Amazon SageMaker. Host the model on a SageMaker real-time endpoint. Configure an Amazon API Gateway API and an AWS Lambda function to handle real-time inference requests that the web application sends. Exclude the items that the user previously purchased from the results before sending the results back to the web application.
Use an Amazon Personalize PERSONALIZED_RANKING recipe to train a model. Create a real-time filter to exclude items that the user previously purchased. Create and deploy a campaign on Amazon Personalize. Use the GetPersonalizedRanking API operation to get the real-time recommendations.
Use an Amazon Personalize USER_PERSONALIZATION recipe to train a model. Create a real-time filter to exclude items that the user previously purchased. Create and deploy a campaign on Amazon Personalize. Use the GetRecommendations API operation to get the real-time recommendations.
Train a neural collaborative filtering model on Amazon SageMaker by using GPU instances. Host the model on a SageMaker real-time endpoint. Configure an Amazon API Gateway API and an AWS Lambda function to handle real-time inference requests that the web application sends. Exclude the items that the user previously purchased from the results before sending the results back to the web application.
Explanations:
This option uses a user-based collaborative filtering algorithm on Amazon SageMaker, which would require significant manual development effort to handle filtering, training, and deployment, making it more complex than using Amazon Personalize.
This option uses the PERSONALIZED_RANKING recipe, which is designed to rank a set of candidate items based on a user’s preferences. However, it doesn’t directly provide recommendations and would not be the optimal solution for this scenario.
This option uses the USER_PERSONALIZATION recipe in Amazon Personalize, which is designed for providing personalized recommendations based on user history. It also allows for real-time filtering, making it the least complex and most efficient solution for the given requirements.
This option requires building a neural collaborative filtering model on Amazon SageMaker, which would involve a lot of custom development and infrastructure management compared to using Amazon Personalize, which is a managed service designed for such use cases.
I arrange that the answer is:
Use an Amazon Personalize USER_PERSONALIZATION recipe to train a model. Create a real-time filter to exclude items that the user previously purchased. Create and deploy a campaign on Amazon Personalize. Use the GetRecommendations API operation to get the real-time recommendations.
If I’m correct, the answer is:
Use an Amazon Personalize USER_PERSONALIZATION recipe to train a model. Create a real-time filter to exclude items that the user previously purchased. Create and deploy a campaign on Amazon Personalize. Use the GetRecommendations API operation to get the real-time recommendations.