Mark Lewis, Haibo Wang, Gary Kochenberger
Annals of Data Science, March 2014, Volume 1, Issue 1, pp 15-23

In this paper, we investigate a natural nonlinear alternative to a standard linear model for CCP and compare the two models on a set of test problems. Our results show that moderate sized instances of CCP can in fact be solved optimally with modern exact methods in modest amounts of time and that the quadratic model generally outperformed its equivalent linear alternative in terms of quickly finding optimal or near optimal solutions.