基于神经网络的山东省机动车保有量预测  

A forecasting on motor vehicles population of Shandong based on BP neural network

作　　者：牟振华[1] 李美玲[1] 赵庆双[2] 

机构地区：[1]山东建筑大学土木工程学院,山东济南250101 [2]聊城职业技术学院,山东聊城252000

出　　处：《山东建筑大学学报》2009年第3期229-232,共4页Journal of Shandong Jianzhu University

摘　　要：分析了预测机动车保有量时的若干影响因素,针对少量影响因素可提高BP神经网络训练速度的特点,提出利用主成分分析法通过SPSS软件来缩减影响因素数量。并根据这些筛选出来的因素,建立神经网络并对其训练达到预定误差后对机动车保有量进行预测。最后通过山东省实例,将BP神经网络预测的数据与趋势外推法预测的数据进行对比,得出BP神经网络法运算速度较快、误差较小的结论。In forecasting the population of motor vehicles, first of all the article mentions some effecting factors. Based on the point work, we filter the main factors ding the main factors and make comparison is made between the that only some factors can accelerate the training in the neural netthrough the software SPSS, then design the BP neural network incluit fit for the forecasting. At last, taking Shandong as trend extrapolation and BP neural network, and it is an example, a concluded that BP neural network is fast and accurate in computation, thus deserving extensive application.

关 键 词：神经网络 保有量 主成分分析 误差分析 

分 类 号：U491[交通运输工程—交通运输规划与管理]