基于改进的PSO算法优化灰色神经网络的话务量预测  

Telephone traffic forecasting based on grey neural network with improved particle swarm optimization algorithm

作　　者：俞秀婷 覃锡忠[1] 贾振红[1] 傅云瑾 曹传玲 常春 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]中国移动通信集团设计院有限公司,北京100080 [3]中国移动通信集团新疆有限公司,新疆乌鲁木齐830063

出　　处：《计算机工程与设计》2015年第11期2902-2905,2947,共5页Computer Engineering and Design

基　　金：中国移动通信集团新疆有限公司研究发展基金项目(XJM2013-2788)

摘　　要：为解决灰色神经网络模型中参数不易确定的问题,提出改进的粒子群(IPSO)算法,寻找灰色神经网络的参数最优解。在寻优过程中引入一个速度阈值,当粒子的飞行速度小于给定阈值时,给粒子施加一个加速动量,重新初始化粒子速度及位置。用训练好的网络预测两个地区的忙时话务量,与未改进的粒子群算法优化灰色神经网络(PSO-GNN)模型、灰色神经网络(GNN)模型、反向传播神经网络(BPNN)模型的预测结果比较,比较结果表明,改进的粒子群算法优化灰色神经网络(IPSO-GNN)模型提高了预测结果的精度。To solve the problem that the parameters in grey neural network are difficult to determine,the improved particle swarm algorithm was employed to search the optimums by the introduction of a threshold of velocity.When the particle velocity was less than the threshold,an accelerated momentum was applied on the particles to reinitialize the particle velocity and position.The proposed approach was used to predict the busy telephone traffic of two regions,and the forecasting results were compared with those of GNN,PSO-GNN and BPNN.The results show the high prediction accuracy of the proposed algorithm.

关 键 词：粒子群算法 灰色神经网络 加速动量 速度阈值 话务量预测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]