Research On Housing Project Cost Estimation Based On Machine Learning Method

When a developer plans to develop a residential area,the project cost will be affected by many factors,and the internal correlation between each influencing factor is relatively large,but the change of the influencing factor will change the data relationship of its internal correlation.Through engineering examples,it is found that the cost of constructing a residential area project mainly depends on its engineering characteristics.The project cost estimation method is not simply a weighted summation between influencing factors.Therefore,based on machine learning algorithm,the establishment of a non-linear mapping relationship between influencing factors and project costs can be used to realize the intelligent estimation of project costs.Provide new methods and technologies.First of all,in this paper,the previous period of residential project cost prediction is the research object.According to the enterprise internal cost management manual,the data characteristics of different project characteristics(factors affecting project cost)are analyzed,and the characteristic composition and data estimation plan are determined.According to the data to be estimated,it is divided into 6 regional objects of Central China,Henan,East China,Beijing-Tianjin Wing,Greater Bay Area and West China,and 15 characteristic factors are used as influencing factors to estimate and analyze the project cost.Based on the improved gray wolf and support vector machine algorithm,this paper proposes an improved gray wolf optimized support vector machine(igwo-svm)method.Then,in order to compare and analyze the advantages and effectiveness of the improved gray wolf optimization support vector machine(IGWO-SVM)in the engineering cost estimation process,first based on the classic neural network algorithm,including BP artificial neural network(BP),extreme learning machine(ELM)and Fuzzy Neural Network(FNN)to establish the fitting relationship between 15 engineering features and engineering cost,and visualize the fitting results.At the same time,it is compared with the classical particle swarm optimization swarm intelligence search method,and uses the correlation The coefficient(R2),root mean square error(RMSE)and relative error(RC)are used to evaluate the estimation effect.Finally,use the radial basis kernel function(RBF)of the SVM to map the 15 low-dimensional feature data to the high-dimensional space for linear estimation,and use IGWO to improve the two important parameters that affect the estimation performance of the SVM,and based on the above R2 RMSE and RC conduct a comparative analysis of the estimation effect.Judging from the evaluation results,the overall project cost estimation effect in Central China,Beijing-Tianjin Wing,and Greater Bay Area is better than Henan,East China and West China.In the classic artificial intelligence method,FNN has achieved the best average estimation effect.The overall average estimated R2 of the 6 regions is 0.9470,the RMSE is 204.6303,and the RC is 4.2283%;the BP is the worst,and the overall average estimated R2 of the 6 regions is 0.8919,RMSE is 329.7094,RC is6.97%.The IGWO-SVM obtained the best estimation effect.The overall average estimated R2 of the 6 regions was 0.9902,the RMSE was 77.9165,and the RC was1.5833%.And based on IGWO-SVM,the project cost of the Xingfuli community of the project example was estimated,and the calculation deviation of 2.39% relative to the actual project cost was obtained.The research results of this paper show that IGWO-SVM can be used as an effective intelligent estimation method to realize the nonlinear fitting between engineering characteristics and engineering costs,and provide new methods and technologies for realizing the intelligent estimation of engineering costs.