How machine learning helps improve your powers of prediction

Machine learning models that analyze enormous amounts of data with seemingly impossible speed and scale enable organizations to make a broad variety of predictions, ranging from when someone might suffer a heart attack, who might get injured in a football game, or when you’ll need to increase your staffing.

These predictions show increasingly high levels of accuracy because the models driving them are often ingesting petabytes of data and billions of parameters – orders of magnitude more than just a few years ago. Predictions are only as good as the data they act upon – and the data and analysis are getting better.

Explosive growth

When Amazon deployed its machine learning service, Amazon SageMaker, about three years ago, most customers would deploy a few dozen models at most, says Bratin Saha, Vice President and General Manager of Machine Learning Services at Amazon. Now, some customers are deploying thousands of models, with billions of parameters, and making hundreds of billions of predictions, he says.

 “As the models become more sophisticated, machine learning is becoming an integral part of doing business across every domain,” says Saha.

For example, in healthcare, Cerner is using ML and AI in its digital health system that acts on real-time information to achieve predictable excellence in clinical, operational, and financial outcomes. The Cerner Machine Learning Ecosystem can make near-real-time predictions about patient care and hospital operations – including hospital capacity and length of individual patient stay – to help clinicians make more informed decisions. Using AWS tools, Cerner gives hospitals the power to integrate their networks, make predictions that empower care teams, and transform teams by democratizing data.

“There is a very carefully curated dataset that needs to be brought to bear in real time for people in healthcare to do their best work and for the clinicians and patients to get their best experience,” says Lisa Gulker, senior director of health system operations at Cerner.

In the sports industry, the National Football League utilizes AI and ML for its Next Gen Stats program to provide fans and teams with probability predictions on individual plays, such as the likelihood of a catch made by a specific receiver from a pass thrown by a specific quarterback. In addition, the NFL is using AWS machine learning services to help make predictions on whether an injury will occur to a player.

“Our goal is to improve player safety by eventually being able to predict and therefore prevent injury,” Jennifer Langton, Senior Vice President of Player Health & Innovation at the NFL, said at AWS re:Invent 2020. “AWS’ AI and machine learning services, combined with the NFL’s data, will speed an entire generation of new insights into player injuries, game rules, equipment, rehabilitation, and recovery.”

Explaining the predictions

As predictions become accurate – and more complex – it’s important to ensure that business leaders understand the reasons behind each prediction, to mitigate bias or to be able to explain the reason behind a prediction. Known as explainability, the concept helps define how models make certain predictions. Amazon’s SageMaker Clarify assists with explainability and bias detection with “feature importance” graphs that help teams explain model predictions and produces reports that can be used to support internal presentations or to identify issues with models that teams can take steps to correct.

“We have customers that have made their machine learning systems more reliable and increased transparency to the stakeholders,” says Saha. “We’ll continue to innovate on this so we can make it easier for our customers to understand why their machine learning models are making the predictions they’re making.”

Learn more about ways to reinvent your business with data.