Research On The Prediction Of Residential Building Approximately-estimated Cost Based On Case-based Reasoning And Wavelet Neural Network

With the gradual increase of social fixed asset investment,China's real estate market has entered a stage of steady development,the cost management of residential buildings has also gradually entered the stage of refined management.However,in the preli minary design stage,on the one hand,due to the limited engineering information,and the cost compilation method is not unified,the estimated cost in this stage generally has the shortcomings of large error and long preparation time,which cannot be well adapted to the needs of economic development.On the other hand,the selected budgetary cost index is difficult to provide a strong reference to the budgetary valuation of this stage,which makes it difficult to provide guidance to the subsequent project cost control and management.In addition,the current research on this aspect is not strongly targeted,and there is no specific division of the cost types of residential buildings in different stages.Therefore,it is of great significance to accurately and efficiently predict the construction cost of residential buildings.With the continuous development of machine learning,artificial intelligence methods have been applied in the field of cost prediction.Taking this as an opportunity,this study,based on the perspective of investors,engineering cost consulting institutions and design units,proposes the forecast model of index system based on case-based reasoning and the intelligent forecast model based on wavelet neural network for the estimated cost of residential buildings in the preliminary design stage.Firstly,on the basis of analyzing the research status at home and abroad,summarizing the evolution process of the prediction model and the shortcomings of the current research,the relevant theories of case-based reasoning and wavelet neural network are expounded.Secondly,based on the principle of combining quantity with price,the factors affecting the estimated cost of residential buildings are summarized,the character input index is quantified,and the rough set software Rosetta is used to reduce the attributes of the factors affecting the initially determined quantity,so as to effectively improve the operation efficiency of the model.Through the network data sharing platform,the project information of 47 residential buildings in eight cities of a province from 2008 to 2017 was obtained to build a case base.Then,MATLAB programming is used to realize the visual and operable graphical user interface of the index system prediction model and the simulation of the intelligent prediction model.The index system prediction model calculates the weight of each index based on the correlation coefficient between the index and the unilateral cost.The case retrieval is realized,and the case correction is made for the combination adjustment of the retrieval results,and the output results and prediction errors of the current case are obtained.The model is mainly applicable to investors and design units.The prediction process can be controlled by changing engineering parameters,and the impact of different engineering characteristics on the cost can be taken into account.It is conducive to investors to control the cost of the subsequent construction drawing design stage and determine the general contracting price of the project,and it is conducive to the design unit to optimize the preliminary design plan.After determining the network structure,initial parameters and the transfer function between the middle layer and the output layer of the intellige nt prediction model,network training is carried out on the data,and the convergence diagram and error curve of the training process are obtained.Finally,the model is simulated and verified with the test data.The model is mainly applicable to the cost consulting institution,which is conducive to its rapid design budget estimation.Finally,the accuracy of the prediction results of the output indexes is analyzed,and the operation results of the two models meet the accuracy requirements of the estimated cost of residential buildings in the primary design stage.The case-based reasoning and wavelet neural network is introduced into the budgetary estimate of the cost of prediction in the process,not only for the project cost in the process of practice make full use of digital information,and realizes the residential building estimate cost forecast information,transforms the ex-post control into the ex-ante control,improves the level of cost management and improves the efficiency of cost forecast,which has certain reference value.