as per AWS document
Example 3: Custom batch recommendations A company uses Custom Recommendations to generate personalized item recommendations for each user in their email marketing campaigns. They ingest 10 GB of data and 5M interactions for training. The company uses a batch inference to generate recommendations for 1 million users. Each recommendation request returns 10 items per user, however the company is only charged for the 1 million requests.
In this case the charges for using Personalize will be:
Data processing and storage charge = 10 GB * $0.05 per GB = $0.50 Solution training charge = 5M interactions ingested for training * $2.00 per 1M interactions = $10.00 Inference charge = 1 million requests * $0.15 per 1,000 real-time recommendation requests = $150.00 Total cost = $0.50 + $10.00 + $150.00 = $160.50
now My Question is Q1 If I get batch recommendations for one user, what cost will I incur? Q2 According to AWS documentation, if I request a set of 10 recommendations for 1 million users, it counts as one million requests. If I instead request 500 recommendations per user, will the price change, or will I receive the 500 recommendations at the same cost? Q3 How to update new events for batch recommendation? Q4 Can we use the putEvents method to add new events to the dataset for batch recommendations? If so, will the next batch of recommendations generated include this new data?
I want to know the cost of AWS Personalize batch recommendations. How is the pricing structured? Can I use batch recommendations for a single user? If so, how many items can I request for a user in one batch recommendation request? Additionally, when I create a solution, will there be separate charges for that? i want to about the betch recommendations price structure
A1: If you only request information for one user then the only aspect that changes is the request count in your calculation. So the inference cost would change but nothing else.
A2: 500 recommendations per user would be the same cost as 10 or 1, though 500 is the maximum number of recommendation requests you can return for one user.
A3/A4: Yes the next batch inference job you run will take the new events into account. However putEvents does not by itself update the trained model, only the models understanding of each user, so if you are running a batch inference job against a dataset that has fundementally changed since you first trained your solution version then you may wish to retrain your model.