Statistics for the sniffles


The onset of winter brings with it a rise in coughing, sneezing and wheezing.

Statisticians from the University of Central Florida receive their first-place award for winning the 2011 Data Mining Shootout during a presentation at the Analytics 2011 conference.

But what impact does winter weather have on hospital stays resulting from the spread of infectious diseases?  And how can one predict which types of storms may result in more hospitalizations?

Those are the questions that teams of college and university students were asked to answer as part of the fifth annual Data Mining Shootout, sponsored by SAS and the Institute for Health and Business Insight (IHBI) at Central Michigan University. The contest gives student and faculty teams an opportunity to solve a real-world data mining problem.

‘Creative solutions’ to real-world problems

A seven-person team from the University of Central Florida captured first place in this year’s contest. Team members were recognized at the Analytics 2011 conference in Orlando, Florida, where they received a $5,000 donation and presented their results to conference attendees. Event organizers also recognized second- and third-place teams, both from Oklahoma State University.

More than 40 teams entered this year’s Data Mining Shootout. Representatives from the IHBI administered the contest by determining the data problem to solve and overseeing the judging.

“These results should really be viewed as creative solutions to a problem that has a lot of real-world elements to it,” said Tracy Irwin Hewitt, Associate Director of the IHBI, in presenting the awards. She also praised the students’ technical expertise, as well as their ability to “translate results into meaningful information that could potentially influence decision making.”

Winter wonderland of data

Teams analyzed weather, population, demographic and hospitalization data to determine the types of storms that had a statistically significant impact on the occurrence of certain infectious diseases. Based on their analysis, teams then built a predictive model that could be used to help healthcare providers plan for fluctuations in patient admittance rates during the course of the year.

Of five different storm types analyzed – cold, flood, thunderstorm, windstorm and winter storms – the team from Central Florida discovered that winter storms had the strongest impact on the highest number of infectious disease categories. They also found that cold storms have a statistically significant impact on a number of respiratory illnesses such as bronchitis, asthma and pneumonia, among others.

“The proposition was that weather, particularly storms, caused more people to spend time indoors, which can increase transmission of infectious diseases and result in hospitalizations,” said Jim Mentele, Senior Research Fellow with the IHBI. “We were really impressed with the submissions, and unfortunately, we could only award prizes to a few.”

Jun Han, team captain of the winning team from Central Florida, said the data mining competition was a practical and relevant exercise in statistical analysis for him and his teammates.

“This project was very challenging,” Han said. “This was just like a real-world problem with complicated data sets. It was an amazing learning opportunity that included data preparation, data mining, modeling and a presentation. I learned so much from the SAS competition. This is a great way to train students.”


About Author

Chad Austin

Communications Specialist

Chad is a member of the Internal Communications team at SAS. He supports the Research and Development division and other technology groups at SAS.

Related Posts

Leave A Reply

Back to Top