1基于pca与narx的市政工程造价组合预测张晓东1，杨圣祥2.pdfVIP

基于PCA 与NARX 的市政工程造价组合预测 1 2 张晓东 ，杨圣祥 （1. 江苏开放大学，江苏 南京 210024 ；2. 英国德蒙福特大学，英国莱斯特） 摘 要：针对工程造价影响因素多，预测困难且周期较长等问题，通过将主成分分析和NARX 神经网络相 结合，提出一种新型基于混合算法的市政工程造价预测方法。利用主成分分析对影响市政工程造价的主要 影响因素进行原始数据处理，消除其相关性，可有效降低数据冗余，也降低神经网络运算易出现局部极小 点概率。以主成分分析数据作为输入，单位面积工程造价作为输出，采用贝叶斯正则化算法构建的NARX 网络神经模型进行市政工程造价预测。算例结果表明，基于PCA 与NARX 进行市政工程造价快速、准确， 证明预测有效可行。 关键词：工程造价预测；主成分分析；神经网络 Forecasting the Cost of Municipal Engineering Based on PCA and NARX Neural Network ZHANG Xiao-dong1 YANG Sheng-xiang2 （1. Jiangsu Open University, Nanjing 210024, China; 2. De Montfort University, Leicester LE1 9BH, UK ） Abstract: There are many factors influencing the cost of municipal engineering. It is usually difficult and time consuming to predict the cost of municipal engineering. This paper proposes a two-phase method to predict the cost of municipal engineering by combining principal component analysis (PCA) and NARX neural network. In the first phase of the proposed method, PCA is used to analyze the correlations among involved factors based on the original data and extract key factors that influence the cost of municipal engineering. In the second phase, taking the key factors extracted by PCA as input and the unit area engineering cost as the output, a NARX neural network model is constructed based on the Bayesian regularization algorithm to predict the cost of municipal engineering. The experimental results show that the proposed method can predict the cost of municipal engineering fast and accurately, which shows that the proposed prediction method is feasible and effective. Key words: engineering cost forecasting; principal component analysis; NARX neural networ