Abstract: Though neural networks have been used in the modeling of logistics center site selection，there are still many problems puzzling engineers such as the complex design procedure of single neural network，the“over-fitting” problem and etc. To solve these problems，a new neural network ensemble model with two-level structure is proposed.In this model，the Bootstrap sample technique is used to generate different data sets for training individual neural networks and particle swarm optimization （PSO） algorithm is used to combine the individuals’ outputs.And the final results are got by combining the ensemble individuals.Simulation results show that the new model is easy to build and promotes the generalization ability.

Key words: logistics center, site selection, neural network ensemble, Bootstrap Sample technique, Particle Swarm Optimization（PSO）

摘要： 针对在建立物流中心选址模型中，单个人工神经网络模型难以确定参数、容易产生“过拟合”等问题，提出一种神经网络二次集成模型，利用Bootstrap可重复采样技术得到不同的训练集来训练产生不同的个体神经网络，采用粒子群优化算法结合个体输出获得神经网络集成，并在此基础上将集成视为个体再次结合。实验结果表明，该模型易于设计且能够提高泛化能力。

关键词: 物流中心, 选址, 神经网络集成, Bootstrap可重复采样技术, 粒子群优化