Test drive: real-time prediction in Java with Amazon Machine Learning

Following up on my two previous Amazon ML articles (batch prediction with Redshift and real-time prediction with the AWS CLI), here's a quickie on implementing real-time prediction with the AWS SDK for Java.

At the time of writing, I'm using SDK version 1.9.31:

• sources on Github: https://github.com/aws/aws-sdk-java
• Javadoc : https://docs.aws.amazon.com/AWSJavaSDK/latest/javadoc/

Here's the Maven dependency:

	<dependency>
	<groupId>com.amazonaws</groupId>
	<artifactId>aws-java-sdk</artifactId>
	<version>1.9.31</version>
	</dependency>

view raw Maven dependency for AWS Java SDK hosted with ❤ by GitHub

The code (source on Github) is pretty self-explanatory and totally similar to the CLI example in the previous article.

• Get the list of models
• Pick one
• Build a prediction request : data record + endpoint
• Fire the request
• Get a result!

	package org.julien.datastuff;
	import com.amazonaws.AmazonClientException;
	import com.amazonaws.services.machinelearning.AmazonMachineLearningClient;
	import com.amazonaws.services.machinelearning.model.DescribeMLModelsResult;
	import com.amazonaws.services.machinelearning.model.MLModel;
	import com.amazonaws.services.machinelearning.model.PredictRequest;
	import com.amazonaws.services.machinelearning.model.PredictResult;
	import com.amazonaws.services.machinelearning.model.RealtimeEndpointInfo;
	public class MLSample {
	public static void main(String[] args) {
	// Create a machine learning client
	AmazonMachineLearningClient client = new AmazonMachineLearningClient();
	// Get list of prediction models
	DescribeMLModelsResult models = client.describeMLModels();
	// Iterate over all models and show basic information about each one
	for (MLModel m : models.getResults()) {
	System.out.println("Model name: " + m.getName());
	System.out.println("Model id: " + m.getMLModelId());
	System.out.println("Model status: " + m.getStatus());
	RealtimeEndpointInfo endpoint = m.getEndpointInfo();
	System.out.println("Endpoint URL: " + endpoint.getEndpointUrl());
	System.out.println("Endpoint status: "
	+ endpoint.getEndpointStatus());
	}
	// Select first model
	MLModel model = models.getResults().get(0);
	// Build a prediction request
	PredictRequest request = new PredictRequest();
	// Select prediction model
	request.setMLModelId(model.getMLModelId());
	// Select realtime endpoint
	request.setPredictEndpoint(model.getEndpointInfo().getEndpointUrl());
	// Build data to be predicted
	request.addRecordEntry("lastname", "Simon")
	.addRecordEntry("firstname", "Julien")
	.addRecordEntry("age", "44").addRecordEntry("gender", "M")
	.addRecordEntry("state", "Texas").addRecordEntry("month", "4")
	.addRecordEntry("day", "106").addRecordEntry("hour", "10")
	.addRecordEntry("minutes", "26").addRecordEntry("items", "5");
	System.out.println("Sending prediction request for: "
	+ request.getRecord());
	// Send prediction request
	PredictResult result;
	try {
	long start = System.currentTimeMillis();
	result = client.predict(request);
	long end = System.currentTimeMillis();
	System.out.println((end - start) + " ms");
	} catch (Exception e) {
	throw new AmazonClientException("Prediction failed", e);
	}
	// Display predicted value
	System.out.println("Predicted value:"
	+ result.getPrediction().getPredictedValue());
	}
	}

view raw MLSample.java hosted with ❤ by GitHub

Here's the output:

	Model name: ML model: Data3.txt
	Model id: ml-noV4tkTZgnO
	Model status: COMPLETED
	Endpoint URL: https://realtime.machinelearning.us-east-1.amazonaws.com
	Endpoint status: READY
	Sending prediction request for: {firstname=Julien, gender=M, month=4, hour=10, minutes=26, state=Texas, day=106, items=5, age=44, lastname=Simon}
	Predicted value:752.1902

view raw MLSample.java output hosted with ❤ by GitHub

Nice and simple. What about performance? My previous article measured a CLI call (aka 'aws machinelearning predict') within us-east-1 at around 300ms, well above the 100ms mentioned in the FAQ.

Believe or not, the Amazon product team got in touch (I'm not worthy, I'm not worthy... and thanks for reading!). They kindly pointed out that this measurement includes much more than the call to the web service and of course, they're right.

Fortunately, the Java program above is finer-grained and allows us to measure only the actual API call. I packaged it, deployed it to a t2.micro EC2 instance running in us-east-1 (same configuration as before) and...

	[ec2-user@ip-172-30-1-10 ~]$ java -jar mlsample.jar
	Model name: ML model: Data3.txt
	Model id: ml-noV4tkTZgnO
	Model status: COMPLETED
	Endpoint URL: https://realtime.machinelearning.us-east-1.amazonaws.com
	Endpoint status: READY
	Sending prediction request for: {minutes=26, items=5, age=44, month=4, state=Texas, gender=M, lastname=Simon, hour=10, day=106, firstname=Julien}
	80 ms
	Predicted value:752.1902

view raw Benchmark hosted with ❤ by GitHub

Average time is around 80ms, which is indeed under the 100ms limit mentioned in the FAQ. There you go :)

I'm definitely looking forward to using this in production. Till next time, keep rockin'.