| At present,the continuous promotion of urbanization has led to the scarcity of urban land resources and serious price increase,and high-rise residential buildings have gradually replaced multi-story residential buildings as the main type of residential construction by virtue of higher plot ratio and lower development cost per unit area.However,the number of floors,large construction area and complex structure of high-rise residential buildings lead to the serious phenomenon of "three super" in the construction process,so limiting the project cost of high-rise residential buildings to a reasonable range is an urgent problem in the construction industry.According to the research of some western scholars,the cost of the initial stage of the project only accounts for about 1% of the total cost of the project,but the impact on the total amount of project cost accounts for 75%.This paper aims to build an engineering investment estimation model according to the theory of intelligent algorithm to achieve rapid investment estimation in the decision-making stage of high-rise residential projects.A lot of data pre-processing work has been done in this paper.The selection of input indicators has a great influence on the estimation effect of the intelligent estimation model,but there are many factors affecting the construction project cost,and considering all of them will lead to inefficient network operation,while the traditional selection of representative indicators based on experience does not have objectivity.In this paper,we review a large amount of literature to establish the principles of index selection,and filter the feature vectors of each unit project of high-rise residential engineering through correlation analysis,and delete the indexes that have no influence on the cost.Finally,the feature vectors are simplified by factor analysis to remove the information overlap between data,and the extracted factors are used as the input vectors of the model to greatly improve the training performance of the network.The BP neural network model,extreme learning machine model(ELM),and support vector machine(SVM)model were constructed in MATLAB software,and the initial weights and thresholds of BP neural network and parameters c and g in support vector machine model were optimized by genetic algorithm to construct GA-BP and GA-SVM models.257 high-rise residential projects in Shanxi Province from 2016-2020 were collected for validation analysis,and the cost of civil engineering,decoration and renovation works,and installation works were estimated separately.The results show that the performance of the genetic algorithm-optimized model is generally improved,the overall fit of the BP model is better,the parameters of the ELM model are easy to determine but sensitive to the sample,and the prediction of the SVM model is the best among the three.Overall,the maximum relative errors of the civil construction model are within 10%,the maximum relative errors of the decoration works and installation works are within 15%,and the maximum relative errors of the total cost of GA-SVM model are controlled within 5%,and the model has certain practicality and can be used as a reference for engineering investment estimation. |
PDF Full Text Request