Study On Prediction Of Construction Cost Based On Particle Swarm Optimization Least Squares Support Vector Machines

The prediction of construction cost is the basis of feasible study,the premise of quota design and the evidence of establishing the base bid in the process of bidding.Accurate and efficient project cost prediction has important theoretical and practical significance for the forward management of construction cost.As the project has a long construction cycle and involves a wide range,the historical data of the project has the characteristics of limited sample size and many attributes.The prediction of construction cost is essentially a kind of small sample learning problem.In recent years,research on engineering cost prediction based on fuzzy mathematics,grey relational grade and artificial neural network is more extensive.But the model design of fuzzy mathematics and grey system theory is too simple,it is difficult to obtain good prediction accuracy,and neural network learning has the disadvantages of large sample requirement,slow convergence speed and poor generalization.Support vector machines provide the best technology platform for learning and processing small sample data,this study combines the particle swarm algorithm to improve the support vector machine,proposes a parameter optimization intelligent learning method based on support vector machine,and apply it to the prediction of construction cost.In this paper,the mathematical model of least squares support vector machine is given based on the basic idea and mathematical principle of support vector machine.Then,on the basis of relevant literature research and expert scoring,combined with the composition of construction cost and its influencing factors,a reasonable prediction index system of construction cost is established.Secondly,using the system cluster analysis to classify the similar sample data,and using principal component analysis to reduce the dimensionality of the attribute index,comprehensive index was not related to each other,so as to reduce the sample complexity and improve the efficiency of learning model.Finally,for defects that parameter setting depends on experience value in the model,with the advantage of particle swarm optimization in parameter optimization,a prediction model based on parameter optimization is proposed to further improve the prediction accuracy and stability.According to the above theoretical results,based on the collection of historical cost data,using SPSS for data preprocessing,and simulation analysis is carried outusing Matlab programming.The applicability and effectiveness of the particle swarm optimization least squares support vector machines model is verified,and it has positive guiding significance for improving the management level of Engineering cost.