It is like time series prediction, but not exactly. Our customer sends patient current vital information to AWS SageMaker inference endpoint to predict the patient future health condition. The patient data is saved to AWS S3 bucket, and we use Athena to build a table from saved data. We did off-line training, and found the vital information change is an import feature to predict patient future health condition. Does anyone know how I can query Athena table to extract a given patient historical vital information, and send both historical and current vital information to the AWS Sagemaker inference endpoint to make prediction?

CodePudding user response：

Do you have a SageMaker real-time endpoint created? One route is using an AWS SDK (Ex: boto3 for Python) and coupling it with a Lambda function to invoke that endpoint. The Lambda function can grab data from S3 or Athena or whatever data source using the SDK and then invoke that endpoint.