层次分析法(AHP)中根据判断矩阵求解排序权重问题本质上为一个使一致性指标最小化的优化问题。针对现有解决方法中的不足,提出一种结合粒子群和模拟退火原理,并且根据AHP的特点引入特征粒子来求解判断矩阵排序权重的算法,同时,针对一致性不满足条件的矩阵或者残缺矩阵,在一致性指标中引入可信度参数,使算法能够动态修正不一致判断矩阵或者残缺矩阵,应用的范围更加广泛。文中对判断矩阵求解排序权重以及一致性检验、模拟退火粒子群算法解决AHP排序问题进行了介绍,并给出了实验数据以及分析。实例结果表明,算法可行且有效,计算结果精度高,稳定性好。
The rank weights computing based on comparison matrix in the an',dytical hierarchy process (AHP) can be attributed to a optimization problem to minimize the consistency index. For the shortcomings of existing solution, an algorithm integrating particle swarm optimization with simulated annealing is presented to compute the rank weights. The eigenvector particle is imported into this algorithm based on the character of AHP. To compute the matrix whose consistency index condition is not satisfied or the incomplete matrix, a reliability parameter is added into the consistency index to correct the inconsistent or incomplete matrix dynamically, which makes the algorithm find wider application. In this paper, using the comparison to compute rank weights, consistency check and how to use the simulated annealing-particle swarm optimization "algorithm to solve computing rank weights in AHP is proposed. The results of case study show that the algorithm is feasible and effective. Its computational results are of both stability and high precision.