Finding nondominated points for multi-objective integer programs and its applications
Department of Industrial Engineering
Middle East Technical University
Finding nondominated points is typically hard for Multi-objective Integer Programs (MOIPs) and the number of nondominated points is large when the objectives are conflicting. It is not practical for a decision maker (DM) to compare all of these points to make a decision. Therefore, it is important to generate nondominated points in a region that is of interest to the DM. In this study, we first present two exact algorithms that efficiently generate all nondominated points for MOIPs. We then develop a variation of those algorithms that generates the true nondominated points in any specified region for MOIPs. To define the preferred region of the DM, we also develop a procedure that first approximates the nondominated set using a hypersurface, finds a preferred hypothetical point on this surface and then defines a preferred region around the hypothetical point. Once the preferred region is defined, all nondominated points in that region are generated. We test the performance of the algorithms on some multi-objective combinatorial optimization problems and demonstrate that the algorithms work well.
This is a joint work with Murat Köksalan.
Banu Lokman is an Assistant Professor in the Industrial Engineering (IE) Department of Middle East Technical University (METU). After receiving her BS degree in 2005, she worked for ASELSAN as a planning engineer during 2005-2006. She received her MS degree in 2007 and her PhD degree in 2011 from the IE department of METU. She was a research assistant in the same department from 2006 to 2011. She joined TED University in 2012. From May 2012 to June 2013, she worked as a visiting researcher in School of Business of Aalto University, on leave from TED University. She has been a member of the faculty in IE Department of METU since September, 2014. Her current research interests are under the umbrella of multi-criteria decision making, in areas such as combinatorial optimization, evolutionary algorithms, and applications.