The quota pricing and bill pricing cannot accurately and efficiently determine the construction cost,while the construction cost index can directly reflect the impact of market price change on the construction cost.Accurate and efficient prediction of construction cost index can provide decision support for all parties of project construction and is the basis for government departments to make relevant policies.In order to accurately and efficiently predict the residential construction cost index,combining the functions and advantages of the center approach GM(1,1)model,the mind evolutionary algorithm(MEA) and the BP neural network prediction model, a MEA-BP neural network prediction model based on the center approach GM(1,1) was constructed.First,the relationships between the single cost indexes of labor,material and machine cost and the comprehensive cost index of residential construction were determined through training samples,and then,the future single cost index of the material machine obtained by the center approach GM(1,1) model was used as the input variable of the MEA-BP neural network to obtain the future comprehensive cost index of residential construction. Using MATLAB 2018a software simulation experiment，the residential construction cost index was predicted,and compared with prediction results of the BPneural network.The comparison between MEA-BP and BPneural network prediction model shows that the MEA-BP neural network prediction model has higher prediction accuracy,which can be used to predict the future residential construction cost index.