This week’s list of data news highlights covers June 14-20 and includes articles about a major machine learning offering from Microsoft and a Wall Street watchdog’s move to the cloud.
1. Data Sharing Platform Allows Cancer Centers to Share Data
Most cancer centers base treatments on clinical trial research, but this research draws from data on only about four percent of patients, since most patients do not participate in clinical trials. Flatiron Health set out to get centers better access to data on the remaining 96 percent of patients. The New York-based company developed a cloud-based storage and analytics platform that enables centers to contribute their data to a central database in exchange for access to the other centers’ data. The database now contains about 550,000 cases, and organizers hope it will provide the sample sizes necessary for new research.
2. Wearable Devices in Health Insurance
Health insurance companies in the United States and Europe could soon begin to integrate data from fitness trackers and other health wearables as the devices become increasingly popular. The devices, which can often track health variables more cheaply and granularly than traditional lab tests, offer a trove of information that could help these companies dynamically price their policies according to patients’ behavior. In particular, a non-invasive blood-glucose monitoring wearable could be useful in the future, since glucose monitoring can give insights into patients’ diets. For now, however, only one in ten Americans owns a fitness tracking device. While several wearable companies have sought to partner with health care providers, little has come of the efforts.
3. Researchers Develop Method to Embed Sensors in Smartphone Screens
Researchers at Montreal’s École Polytechnique and materials manufacturer Corning published a paper this week detailing a method for embedding sensors in the glass that makes up smartphone screens. At present, phone designers must place certain sensors, such as heart rate monitors or fingerprint readers, in awkward places because the screen makes it impossible to use wired sensors on the front side of the device. The new method, however, uses light to transmit data from the user through the glass and into the device for processing. The researchers demonstrated their method with a basic temperature sensor.
4. Microsoft Reveals Cloud Service for Machine Learning
Microsoft revealed a major new predictive analytics service especially tailored to its Azure cloud platform this week. The service, called Microsoft Azure Machine Learning, will be available to the public in July. Creators hope the service will dramatically lower the barriers to entry for predictive analytics, which has traditionally required extensive data science experience. According to Microsoft, Azure Machine Learning will be easier for less experienced developers to use than ad-hoc solutions or other dedicated machine learning software.
5. Nike’s Data-Driven World Cup Uniforms
To help players cope with Brazil’s unpredictable weather during the World Cup, Nike designers working on the U.S. team’s uniforms first looked to data. They collected information on sweat patterns in a simulated climate chamber to determine where players sweat the most and then replaced the ordinary fabric in those areas with a breathable mesh. They also tested different ways to allow air to flow through the jersey, settling on laser-cut ventilation holes on the sides of the jerseys.
6. Bank of America will Launch Database to Spot Bad Payday Lenders
Bank of America will soon deploy a new database to flag payday lenders that might be violating state laws. The database, developed by the New York Department of Financial Services, was created to help enforce the state’s ban on high rate loans. Although Bank of America is the first bank to sign up for the system, the system will enable the bank to notify other banks of lenders who have made illegal loans in New York.
7. Researchers Mine Patent Data for Trends
Researchers at Oxford University have mined the U.S. Patent and Trademark Office’s (USPTO) rich dataset of records reaching back to 1790 in order to study the changing dynamics of invention over the past two centuries. One question they wanted to answer was whether invention is a combinatorial process, in which inventions come mainly as the product of combinations of other inventions or whether inventions tended to happen in more isolated circumstances. Using the USPTO’s elaborate system of patent codes, the researchers found that inventions relied on combinations of previous technologies around 40 percent of the time, while around 60 percent introduce entirely new technologies.
8. Netflix Moves Beyond the Recommendation Algorithm
Although Netflix’s best known use of data is in its sophisticated movie recommendation algorithm, the company is using data in a number of other interesting ways. For one, they are studying how changes in various quality metrics, such as bitrate and rebuffer rates, affect users’ behavior. Netflix is also deploying natural language processing and text analytics to make closed captions and subtitles more accurate, a problem that has become particularly acute as the company has expanded globally.
9. Insurer Uses Data to Improve Patient Prescription Adherence
The Community Health Plan of Washington (CHPW), a health insurer, has piloted using patient data to improve patients’ adherence to prescriptions. The insurer partnered with Health Integrated, a behavior change and informatics firm, to pull data from multiple sources, identify the patients most at risk of deviating from their prescriptions, and recommend the best courses of action. The data includes information collected from caseworkers, veterans’ officials, nutrition assistance programs, and other sources. CHPW found that the program has shown improvements in prescription adherence and patient satisfaction.
10. FINRA Seeing Results from Switch to Cloud Storage and Analytics
The Financial Industry Regulatory Authority (FINRA) has begun transitioning to the cloud and incorporating analytics to become a more efficient Wall Street watchdog. The agency, which will now store and process data from its surveillance and other oversight duties on a remote server, projects that it will save $10 million to $20 million annually after a 30-month rollout, in part due to the cloud provider’s pay-per-use policies. In addition, the move will allow the agency to efficiently process the 25 billion daily market events it currently tracks increases. FINRA officials have already described dramatic improvements in processing speeds.
Photo: Flickr User Robert Scoble