Asking the right questions – Shahram Rahimi, professor and chair of the Department of Computer Science at Southern Illinois University Carbondale, works with students at the Dunn-Richmond Economic Development Center at SIU. Rahimi’s group is working on software that can analyze so-called “big data” using a two-phase, $220,000 grant from a major health care corporation. Rahimi and his students are building and testing software that could lead to huge improvements in patient care and satisfaction in that sector. With a few tweaks, it also might be used in many other industries and government functions that rely on interpreting big data sets. (Photo by Steve Buhman)

July 10, 2015

Research focuses on improving ‘big data’ analysis

by Tim Crosby

CARBONDALE, Ill. – If you take a mountain of data and run it through a common computer, while asking it a specific question about the data, it can probably find the answer. 

But what if the question you asked isn’t the right question? What if the mountain of data actually contains information that is far more important than the question you asked it? And what if the computer was smart enough to look at the data and, on its own, tell you what it sees? 

A computer science researcher at Southern Illinois University Carbondale is working on software that just may accomplish such a feat. Working in the world of so-called “big data,” Professor Shahram Rahimi said advances in computer hardware and artificial intelligence are sending researchers in new directions and prompting the search for new algorithms that could lead to revolutionary improvements in how we use massive data sets. 

With a three-year, two-phase $220,000 grant from a major health care corporation, Rahimi and his students are building and testing software that could lead to huge improvements in patient care and satisfaction in that sector. With a few tweaks, it also might be used in many other industries and government functions that rely on interpreting big data sets. 

Advances in computer storage, processing speed and the increasing number of affordable data-gathering devices, such as sensors and even cell phones, have made gathering and storing large amounts of data easier. 

But gathering data isn’t of much use if people can’t distill the information it contains. That part of the equation, the analysis, has not kept pace with the other side. 

Despite the health care sector’s major advances in technology, it has been somewhat slow to embrace data analysis, Rahimi said. But getting a handle on what any big data set is telling you can be a major challenge, he acknowledged. 

“Often they are dealing with large amounts of data – big data – and the different data sets can affect each other as well,” he said. “But the health care sector have not been using technology for things like decision support, or providing guidelines on how to do a better job handling patients, or how they can improve satisfaction, improve quality of care and minimize mistakes and adjust practices for better performance. We want to make it easier to look at this entire picture, while looking for ways to improve things from diagnosis to treatment. 

“The approach we’re working on would look at all this data: the situation, the people, the time things occur, and then analyze it all,” he said. 

A pile of data is just that – a morass of information without context or meaning. Humans traditionally have teased meaning out of it by writing software containing certain algorithms that can sift the data according to a specific instruction. 

But advances in artificial intelligence and concepts such as “fuzzy logic” are making it possible for computers to take on beneficial human cognitive functions while augmenting them with pure computational horsepower. 

“If I give you some data and say ‘look at this and tell me what the average satisfaction level that people report with this doctor,’ you can do that pretty easily and so can a computer today,” Rahimi explained. “But if I give you the data and say ‘tell me something about this data – just tell me what you see. Well, what do you mean? There’s so much data. What do you want to know? And I say, I don’t know, just tell me something about it.’ So if the objective is not there, the target is not there, it’s a much more complex operation. 

“So this system we are working on would be able to tell you what it sees itself without you giving it a target,” he said. “It will look at the data and tell you what it thinks is important based on patterns it sees and other factors. It basically comes up with question and answer.” 

Humans can look at two or three dimensions usually, with some very talented individuals being able to digest and compare four dimensions. But a patient satisfaction survey, for instance, might contain dozens of questions, each one representing a dimension. 

“A 40-question survey is like having 40 dimensions. How can we possibly analyze that? Well, computers do that very well.” 

What they don’t do that well – yet – is use human concepts such as getting close, rather than exact, the fuzzy logic concept. Nor do computers currently do well recognizing patterns without parameters provided by a human operator. 

Advanced software, such as the type Rahimi and his students are developing, would allow for true artificially intelligent analysis of areas such as emergency room practices, patient satisfaction and outcomes and decision-making. In the end, it could communicate with humans via mapping and other visualization software to highlight areas where there are problems. 

“It could tell you, for instance, that a certain doctor working at night with children who have abdominal pain is not really the best situation,” Rahimi said. “It can figure out what the data actually says, not what you ask it to figure out.” 

And with a few tweaks to the innovative algorithms that power such software, it could be used in a huge number of applications and sectors. 

For now, the work goes on in a room at the Dunn-Richmond Economic Development Center at SIU, where a group of graduate students supervised by Rahimi develop new algorithms and test them on data sets. Their work is paid for by the grant provider, Envision Health, which operates some 700 hospitals across the country. Rahimi hopes the money acts as a seed that would eventually establish a center at SIU to develop new big data analysis software. 

Soon Rahimi will begin testing the results on SIU’s new high-performance computer, a mid-range super computer that came online earlier this year, providing researchers with superior computational power and data storage capacity. 

“Without that new capability, I wouldn’t have even have tried to get a grant for this kind of work,” Rahimi said. “It’s a very important development for research here at SIU.”