基于遗传神经网络的工程造价估算方法研究 |
Research on Model of Engineering Cost Estimation Based on Genetic Neural Network |
投稿时间:2010-05-17 |
中文关键词:工程造价 BP 神经网络 遗传算法 神经网络集成 |
英文关键词:pro ject co st BP neural netw o rk g enetic algo rithm neural netw ork ensemble |
基金项目:国家自然科学基金项目( 70373032) |
|
摘要点击次数: 1329 |
全文下载次数: 886 |
中文摘要: |
人工神经网络已经成功的运用到工程造价估算方法研究中, 高度的鲁棒性和容错能力使它优于多元线性判别分析( MDA) 、逻辑回归等方法( Log ist ic Regr ession) , 针对传统的BP 神经网络在工程造价估算方法中存在收敛速度慢和容易陷入局部最小值等问题, 提出遗传神经网络的估算方法。将遗传算法和神经网络结合, 充分利用两者的优点, 使新算法既有遗传算法的全局随机搜索能力, 又有神经网络的学习能力和鲁棒性。利用遗传算法的全局搜索能力, 针对传统误差反向传播算法的不足, 采用染色体编码对神经网络的权值和阈值等主要参数进行优化, 通过仿真试验验证其稳定性和有效性, 表明该算法在工程造价估算方法中具备较高的实用性。 |
英文摘要: |
Art if icial neur al netw ork has been successfully applied to pro ject cost est imat io n metho ds study . It s hig h degree of ro bustness and fault tolerance makes it superio r to mult iple linear discriminant analysis ( MDA) , log ist ic regr ession and other methods. This paper, in view of slow converg ence and local minimum value of the t raditio nal BP neural netw ork in project co st est imatio n metho ds, proposes genet ic neur al netw ork est imation method. By combing g enetic algor ithms and neural netw ork and fully ut ilizing the adv antages of both, the new algor ithm no t only has the random search capability of g enetic alg orithm, but also has the learning ability and robustness o f neural netw o rk.With the global search capabilities o f g enetic algor ithm, considering the draw backs of t radit ional erro r back??pr opag at ion alg orithm, using chromosome coding, weight s and thr esholds and o ther key parame??ters o f neural netw o rk ar e o pt imized. The stability and ef fect iveness of that alg orithm is v erif ied thro ug h simulat ion test s, w hich show s this method has a hig h pract icality in cost est imat ion.???????? |
景晨光,段晓晨.基于遗传神经网络的工程造价估算方法研究[J].石家庄铁道大学学报:社会科学版,2010(4):11-. |
查看全文 查看/发表评论 下载PDF阅读器 |
|
关闭 |
|
|
|
|
|