Tabu Search
Staff scheduling at the United States postal service
Computers and Operations Research
Hi-index | 0.00 |
This paper presents a methodology to generate diverse solutions for linear programming relaxation and its application to solve integer programs using a neighborhood search. The algorithm generates multiple linear programming solutions of maximal difference and uses them as targets for a neighborhood search to locate high-quality integer solutions. The multiple diverse solutions provide a good coverage of the solution landscape and the neighborhood search reduces the computational effort in searching for good integer solutions. The algorithm is seeded into a genetic algorithm on benchmark knapsack problem and the results have shown that the algorithm is computationally effective, providing high-quality solutions at faster convergence rates than state-of-the-art commercial integer program solvers.