投影向量机在工程单方造价预测中的应用  

Quick estimation of engineering construction cost based on projection vector machine

作　　者：邓万宇[1] 

机构地区：[1]西安邮电学院计算机学院,陕西西安710121

出　　处：《西安邮电学院学报》2011年第5期47-53,共7页Journal of Xi'an Institute of Posts and Telecommunications

基　　金：国家自然科学基金资助项目(61100166;60825202;60803079;60633020);国家高技术研究发展基金资助项目(2008AA01Z131);国家科技支撑计划基金资助项目(2006BAK11B02;2006BAJ07B06)

摘　　要：针对目前工程单方造价预测速度慢、时效性差的问题,提出一种基于投影向量机的快速预测方法,该方法通过对数据降维获得神经网络的输入权和隐含层节点的数目,通过广义逆获得神经网络的输出权。因为整个训练过程不需要迭代调整,所以具有高效的训练速度。在真实单方造价数据上的实验显示,新的预测方法比传统方法速度提高18倍,且精度提高40%,时效性与预测精度都得到显著提高,能够更好地满足造价预测的需要。To solve the time -consuming problem of current engineering cost estimation algorithms, an estimation algorithm based on Projection Vector Machine is proposed. By the new algorithm, input weights and the number of hidden neurons are determined by dimension reduction on input dataset, while the output weights are computed by least square. The new algorithm is fast and efficient. Experimental results on the real data show that, it＇s of a faster training speed by 18 times and a higher accuracy by 1.4 times, compared with the traditional ones.

关 键 词：工程造价预测 投影向量机 自适应调整 

分 类 号：TU723.3[建筑科学—建筑技术科学]