| The cost prediction of construction projects is an important prerequisite for investment estimation by investors during the feasibility study stage,which can directly affect the income acquisition and value appreciation of engineering projects.However,the project definition used in cost prediction is generally not precise enough,and the project characteristics are not yet clear,which reduces the predictive effectiveness of traditional research methods.Therefore,on the premise of fully integrating the engineering characteristics and element attributes of construction projects,this article introduces data mining methods to analyze sample engineering data,in order to establish the correlation between cost prediction indicators(input indicators)and engineering cost item indicators(output indicators),And a machine learning algorithm with Support Vector Machine(SVM)as the core is used to establish a cost prediction model and a deviation risk assessment model,thereby achieving effective prediction of construction project cost and evaluation of deviation risk.According to the structural design of the paper,the main research content includes the following three aspects:This article summarizes the commonly used methods for predicting and evaluating construction project costs through a literature search and data analysis system,and determines the machine learning algorithm based on SVM as the core data mining method of this article.Subsequently,through the analysis and organization of a large number of literature,combined with the engineering characteristics and element attributes of construction projects,this article systematically analyzes the predictive indicators of construction project costs,defines the connotation and quantitative standards of each indicator,and divides them into categories based on input indicators and output indicators.This article is based on the method concept of data mining,and constructs an SVM prediction model and a Principal Component Analysis SVM(PCA-SVM)model after dimensionality reduction of indicators.126 engineering samples are tested by dividing the training and testing sets.At the same time,traditional multiple regression analysis methods were introduced to construct MRA(Multiple Regression Analysis)prediction models for testing.Test the prediction accuracy of single indicators and the prediction effect of cost items based on the test results of three types of prediction models.Finally,this article introduces the Least Squares Support Vector Machine(LSSVM)algorithm to evaluate the deviation risk of construction project costs,and selects an adaptive kernel function to construct an LSSVM deviation risk assessment model.At the same time,the basic process of risk assessment for construction project cost deviation was designed,and the basic content of the five stages of the basic process was sorted out.Then,a reasonable cost interval was obtained by setting the fitting residual of the LSSVM deviation risk assessment model to follow a normal distribution.Then,through empirical testing of 24 construction engineering samples,the optimal predictive value of the engineering cost(Y1)indicator is evaluated,the deviation of the predicted cost is judged,and corresponding deviation risk disposal strategies are proposed.According to the research results of this article,it can be concluded that the SVM prediction model and PCA-SVM prediction model established in this article can effectively predict the cost of construction projects,and are compared with the MRA model supported by multiple regression analysis methods... |
