厦门市工程估价的RBF神经网络预测模型

Prediction Model of the Project Cost Estimation Based on RBF Neural Network

选取55个厦门市典型工程造价指标,利用SPSS统计分析软件对工程特征和训练样本进行相关性分析、归类合并,得出11个工程特征作为平米造价的主要影响因素.以径向基函数(RBF)神经网络原理为基础,建立工程造价估算模型,通过试验,选择net=newrb(P,T,0.01,1.0)建立RBF网络,用y=sim(net1,P)对样本进行训练测试.实证分析结果显示:该模型具有计算快捷简便的优势,估算误差在允许范围内,可用于实际工程造价的辅助估算.

55 typical engineering cost indexes in Xiamen city were selected and analyzed, to obtain the correlation of engi- neering features and training samples by SPSS statistical analysis software. Based on radial basis function （RBF） neural network theory, the project cost estimation model is established. Selecting the net= newrb （P, T,0. 01,1.0） to establish RBF network through the test, using y=sim （net1 ,P） to train and test, the calculation results show that the model has the advantage of convenient calculation, the estimation error is small and allowable, and the model is worth in the project prediction estimation.