Data Mining Driven Neighborhood Search

INFORMS Journal on Computing, Volume 24 Issue 2, Spring 2012 Pages 210-227

University of Alberta School of Business Research Paper No. 2013-1098

30 Pages Posted: 2 Jul 2013 Last revised: 2 Jul 2014

See all articles by Michele Samorani

Michele Samorani

Santa Clara University - Information Systems and Analytics

Manuel Laguna

University of Colorado at Boulder - Leeds School of Business

Date Written: December 2, 2010

Abstract

Metaheuristic approaches based on neighborhood search escape local optimality by applying predefined rules and constraints, such as tabu restrictions (in tabu search), acceptance criteria (in simulated annealing), and shaking (in VNS). We propose a general approach that attempts to learn (offline) the guiding constraints that, when applied online, will result in effective escape directions from local optima. Given a class of problems, the learning process is performed offline and the results are applied to constrained neighborhood searches to guide the solution process out of local optimality. Computational results on the Constrained Task Allocation Problem (CTAP) show that adding these guiding constraints to a simple tabu search improves the quality of the solutions found, making the overall method competitive with state-of-the-art methods for this class of problems. We also present a second set of tests on the matrix bandwidth minimization problem.

Suggested Citation

Samorani, Michele and Laguna, Manuel, Data Mining Driven Neighborhood Search (December 2, 2010). INFORMS Journal on Computing, Volume 24 Issue 2, Spring 2012 Pages 210-227 ; University of Alberta School of Business Research Paper No. 2013-1098. Available at SSRN: https://ssrn.com/abstract=2284385

Michele Samorani (Contact Author)

Santa Clara University - Information Systems and Analytics ( email )

500, El Camino Real
Santa Clara, CA 95053-0382
United States

Manuel Laguna

University of Colorado at Boulder - Leeds School of Business ( email )

Boulder, CO 80309-0419
United States

Register to save articles to
your library

Register

Paper statistics

Downloads
20
Abstract Views
594
PlumX Metrics