采用BP和RBF神经网络的厦门市工程造价预测模型

Project Cost Prediction Model Based on BP and RBP Neural Networks in Xiamen City

收集55个厦门市典型工程造价指标,利用SPSS软件对数据进行预处理,选取11个工程特征作为造价的主要影响因素,分别建立基于多层前馈(BP)和径向基函数(RBF)神经网络的工程估价模型.从55个案例中随机抽取10个作为预测样本,剩下的45个作为训练样本,进行BP,RBF神经网络预测模型的训练和测试.结果表明:通过参数优选的RBF神经网络工程造价预测模型,预测误差在5%以内,网络泛化能力更优越,可用于实际工程造价的辅助估算.

By collecting 55 typical engineering cost indexes in Xiamen City and selecting 11 engineering feature cost per square meter as the main influencing factors, with the help of software SPSS, the neural network engineering cost estima- tion model was established based on back propagation （BP） and radial basis function （RBF）. 10 cases in 55 cases were drew randomly as predicted sample, and the left 45 cases were taken as training sample, BP and RBF neural network pre- diction model were trained and tested. The results showed that the prediction error of RBF neural network through pa- rameter optimization for project cost prediction model is within 5 %, the network＇s generalization ability is benign, so the model can be used for the actual project cost auxiliary estimation.