基于CSO-ELM模型的工程施工成本预测研究

Research on Project Construction Cost Prediction Based on CSO-ELM Model

鉴于影响工程施工成本因素之间复杂的非线性关系，进行准确的工程施工成本预测有一定难度，提出鸡群算法（CSO）和极限学习机（ELM）结合的CSO—ELM工程施工成本预测模型．首先利用CSO对ELM模型的输入权值及偏置值进行全局搜索寻优，得到最佳参数；然后将该参数代入ELM模型中建立CSO—ELM工程施工成本预测模型；最后以11个气膜钢筋混凝土储仓工程为例，验证该模型的科学性．结果表明：CSO优化ELM的输入权值与偏置值是有效的；与传统ELM、BP神经网络模型相比，CSO—ELM模型具有更高的预测精度及效率，为工程施工成本预测提供了一个有效的方法．

In view of the complex nonlinear relationship among the construction cost factors, there has certain difficulty to predict project construction cost the engineering accurately, so the project construction cost prediction model CSO-ELM combined by the Chicken Swarm Optimization （CSO） and the Extreme Learning Machine （ELM） is proposed. Firstly, the input weight and offset value of the ELM model are optimized by CSO, then the optimal parameters are obtained. Secondly, the parameters are substituted into the ELM model to establish the CSO-ELM project construction cost prediction model. Finally, the model is verified to be scientific by 11 concrete dome using inflated forms projects. The results show that it is effective to optimize the input weight and bias value of ELM using CSO, and compared with the traditional ELM and BP neural network model, the CSO-ELM model has higher prediction accuracy and efficiency, which provides an effective method for prediction for project construction cost prediction.