Morozov Earns Moore Award for Dissertation


Sergey Morozov and his adviserEECS Ph.D. student Sergey Morozov has won the Department’s Richard K. & Wilma S. Moore Award for Best Dissertation. Morozov’s research improves the accuracy of recommender systems, which offer personalized suggestions for new books, movies, and other products. This kind of tailoring—based on user feedback and history and purchases of others with similar interests—is extremely profitable; Netflix and Amazon reported over 60 and 35 percent of their sales, respectively, were generated from recommendations.

The huge economic potential of reliable recommender systems is driving development of more sophisticated, faster algorithms. Morozov is working on new collaborative filtering techniques to uncover patterns of behavior and other information that will hone suggestions. Unlike many existing approaches, his recommender system uses feedback from users as well as items. It employs instance selection techniques that pick the best users/items as the basis of the recommendation.

For instance, Morozov says that finding two people who agree on every movie is unlikely. A more realistic method is to first determine which movies are most relevant to the current recommendation and then look for people with similar interests in this sub-domain. Current systems recommend a comedy based on other comedies, without ever inspecting the genre of each movie.

“I believe that data mining has incredible future potential, especially on the Internet. The Moore award recognizes the numerous applications of my work and its importance,” said Morozov, who just completed his first year as an assistant professor at the University of Detroit Mercy (UDM).  While he successfully defended his dissertation last July, Morozov did not officially graduate until this spring.

Morozov says that his research, under the direction of EECS Professor Hossein Saiedian, was a critical piece of his Ph.D. work, as was his teaching. He taught Foundations of Information Technology (EECS 128) and Introduction to Computing (EECS 138). Morozov said his biggest challenge was tailoring the technical aspects of computer science for non-engineering students.

“Programming requires a certain degree of perfection: a missing comma or a misplaced bracket could break an entire system. Teaching such a scrupulous process was difficult at first, but by the end of my fourth semester, I was quite comfortable,” said Morozov, who credits EECS Associate Professor Nancy Kinnersley for giving him the opportunity to experience all aspects of classroom teaching.

“Sergey was an outstanding graduate student,” said his adviser, Dr. Saiedian. “He was an excellent researcher and, at the same time, an excellent teacher. He has contributed to multiple journal and conference papers based on his dissertation research. He genuinely loved teaching and devoted considerable time to his TA [teaching assistant] position at KU. While he had multiple employment opportunities, he chose one that balanced research and teaching.”

Morozov spent six years at KU, earning a master’s degree in 2007 and his doctorate this spring. He has fond memories of his time in Lawrence and encourages students to get to know their classmates and the town. Morozov says study groups that meet regularly can be a lifesaver in a tough class, such as Dr. Saiedian’s “Software Architecture,” which is a course Morozov now teaches. He credits EECS Professor for showing him the power of data mining and EECS Professor Arvin Agah for providing helpful, timely interview advice.

“Lawrence is beautiful, everything is open late, and you can walk everywhere,” said Morozov.  “I really miss getting ice cream at Sylas and Maddy's and walking around Mass St. I could study at the Java Break all night and get cereal at 3 a.m. The KU campus is fantastic.”

At UDM, Morozov teaches undergraduate and graduate courses in software engineering. He recently received an internal grant support for his recommender systems research. He and colleagues are working on a paper for the upcoming Association for Computing Machinery (ACM) Recommender System conference.