A study led by a Binghamton University Ph.D. candidate is looking at a new, more efficient way to predict flight delays, potentially providing airlines with the ability to better inform their passengers.

The project was led by postdoctoral associate Sina Khanmohammadi, who received his Ph.D. in systems science in October. He was assisted by Salih Tutun, a fourth-year Ph.D. candidate in industrial and systems engineering, and Yunus Kucuk, a second-year graduate student in computer science.

Currently, flight delays are predicted using an artificial neural network, a computer program that is designed to simulate the biological learning process of the brain. This works by using existing data from previous flights, and analyzing different variables of that data to look for patterns and estimate an outcome.

The use of these programs is flawed, however, due to the absence of an ability to handle qualitative variables. Factors such as weather and security have no intrinsic ordering to them, and cannot be categorically measured in the same way that data from previous flights can be. The BU researchers looked at ways in which they could “layer” the inputs, meaning they could utilize variables from multiple sources instead of just one.

Using the records of inbound flights at John F. Kennedy International Airport from January 2012 as a case study, they took 15 different variables — including day of the week and time of arrival — and implemented them into their artificial neural network model. This method allowed for multiple variables to be accounted for, limiting error and increasing the accuracy of the programs.

After testing the model with hypothetical flights, they were able to predict delays with approximately 20 percent more accuracy than the current models. The program also arrived at the conclusions of flight delays faster, taking about 40 percent less time than traditional models. While the research has not yet been implemented, Kucuk said that when it is, it will benefit both airlines and their passengers.

“With more accurate flight delay information, managers can make the best arrangements and decide the best schedule for flights to reduce delays,” Kucuk said. “Also, customers can be notified about delays faster. That would help the airline companies gaining respect from their customers.”

According to Khanmohammadi, the research took a problem that was traditionally looked at from one perspective and tackled it through collaboration between members of different departments in addition to the industrial and systems engineering department.

“I believe our research is an indication of the fact that here at Binghamton University, and especially in the industrial and systems engineering department, we are looking at practical and everyday problems from an interdisciplinary perspective and trying to solve these problems using some creative approach,” Khanmohammadi said.

Khanmohammadi now works at Washington University in St. Louis as a postdoctoral associate in the electrical and systems engineering and neurology departments. Since publication in Procedia Computer Science, a scientific journal, he and his team have received widespread coverage from sources such as the New York Daily News and the Economic Times.

“The response to this research was outstanding and at the same time totally unexpected,” Khanmohammadi said. “The reflection of such recognition is the fact that prediction of flight delays is still an unsolved problem and there is a great potential for research in this area.”