设备定位问题局部搜索算法的实验

Test of local search algorithms for uncapacitated facility location problems

讨论设备问题的局部搜索近似算法及其在实际计算中表现出的新性质。主要讨论局部搜索算法中初始解的产生方法,设备价值与服务价值大小对算法求解性能的影响。实验表明:约有99%以上的实例可直接利用局部搜索算法求得最优解;贪心算法产生初始解的局部搜索算法求解时间明显短于随机算法产生初始解的方法,但两者求解质量相当;设备价值和服务价值数值范围越大,局部搜索算法越容易求得最优解。

This paper discusses approximation local search algorithms for Uncapacitated Facility Location Problems（UFLP） and its new property in actual computation.This paper mainly discusses different generation methods of the initial solution,the influence of facility cost and service cost and the improvement of local search algorithm.The testing data shows：There are about 99% instances for which local search algorithm can be used to directly compute their optimal solutions;the running time of the algorithm using initial solutions of greedy algorithm is shorter than that of algorithm using initial solutions of randomized algorithm,but the quality of solutions is almost the same;the bigger the facility cost and the service cost are,the easier local search algorithm gets the optimal solutions.