On Investment Estimation Method of Comprehensive Pipe Gallery Based on PCA-PSO-LSSVM
In order to effectively solve the problems that the existing comprehensive pipeline corridor investment estimation methods are not high in prediction accuracy,and the prediction accuracy is easily affected by sample size,redundancy or lack of feature parameters,a method of Principal Component Analysis(PCA)and Particle Swarm Optimization(PSO)optimized least square support vector machine(LSSVM)combined with a comprehensive pipeline gallery investment estimation forecast model.It uses PCA to reduce the dimensionality of the characteristic parameters that affect the investment estimation of the comprehensive pipe gallery,eliminate noise or redundant data,use the principal component with a larger contribution rate as the input vector of LSSVM,and use the cost per kilometer of the comprehensive pipe gallery as the output vector of LSSVM.It uses PSO to optimize LSSVM's kernel function parameter σ and penalty factor parameter C,establishes a comprehensive pipeline corridor investment estimation prediction model based on PCA-PSO-LSSVM,and predicts the test set samples.The prediction results show that the average relative error of the PCA-PSO-LSSVM model is 3.28%,which meets the requirement of investment estimation prediction error(±10%)in the investment decision-making stage,and is compatible with the PCA-LSSVMmodel,PSO-LSSVMmodel,and GA-BP model.Compared with the GA-SVM model,the prediction accuracy is improved by 67.29%,70.52%,48.13%and 38.60%respectively.The PCA-PSO-LSSVM model has high prediction accuracy and excellent generalization performance.It can be used as an effective prediction method for comprehensive pipeline gallery investment estimation.
NETL
NSTL
万方数据
维普
