工程研究——跨学科视野中的工程2026，Vol.18Issue(1)：70-84,15.DOI:10.3724/j.issn.1674-4969.20240064
基于混合策略ISSA-XGBoost的高速公路工程造价预测研究
Research on Expressway Engineering Cost Prediction Based on Hybrid ISSA-XGBoost
摘要
Abstract
Expressway cost prediction is an important means of cost control in the preliminary stage of expressway construction projects.In this paper,for the characteristics of high dimensional and nonlinear relationship between the small sample data and the engineering cost feature indicators in engineering practice,the XGBoost algorithm with high training efficiency and good prediction performance is used to construct the prediction model,and at the same time,there are many hyperparameters of XGBoost,and the good or bad training result of the model depends on the selection of the hyperparameters,so this paper uses the Sparrow Search Algorithm(SSA)to search for the optimum of XGBoost hyperparameters. Considering that SSA is easy to fall into the local optimum,a strategy that hybridises the sine-cosine algorithm and Lévy's flight is used to improve the basic SSA(ISSA),improve the global search ability of the algorithm and accelerate the convergence speed. In the data collection part,this paper fully considers the influence of project indicators and macroeconomic indicators on the cost of expressways,and finally determines eight cost impact indicators as input features,namely,route length,subgrade width,quantity of earthworks per unit,the ratio of bridges and tunnels,number of interchanges,permanent land occupation per unit,CPI,and material price. In the model establishing part,this paper firstly compares the iterative convergence curves of the three optimisation algorithms of PSO,SSA and ISSA,and it can be seen that the ISSA algorithm is better than SSA and PSO algorithm in both convergence speed and convergence accuracy,which verifies the robustness and effectiveness of the improved sparrow search algorithm. At the same time,the improved sparrow search algorithm is used to optimise the three models of XGBoost,RF and SVM for comparison.The results show that compared with PSO-XGBoost,ISSA-RF,SSA-XGBoost and ISSA-SVM models,the ISSA-XGBoost model shows better antioverfitting performance.Meanwhile,according to the results of each evaluation metric,the ISSA-XGBoost model performs optimally compared with the other four models in the three metrics of MAER,MAE and RMSE,which indicates that the ISSA-XGBoost model is significantly better than the other models in dealing with the overfitting problem and the generalised regression prediction,and that the model has a good adaptability and interpretability,and at the same time,the improved sparrow algorithm improves the training efficiency and prediction accuracy of the traditional XGBoost model. According to the results of the importance of the model features,the three features of the ratio of bridges and tunnels,route length and quantity of earthworks per unit have the most significant influence on the cost of expressways,while the two macroeconomic indicators CPI and price index also have a certain impact on the cost. In order to verify the applicability of the model in actual projects,this study selected three engineering samples outside the input sample set for prediction analysis,and the prediction error rates of the ISSA-XGBoost model for these three samples are-8.87%,-13.44%,and-7.94%,respectively,which meet the requirements of the pre-investment estimation. Through the random sample verification,the ISSA-XGBoost model proposed in this paper can effectively predict the cost of expressways in various regions of the country,which is broad and universal,and the prediction model can be migrated to other fields,providing a certain basis for the construction of efficient and perfect expressway cost management system.关键词
高速公路/造价预测/Lévy飞行/改进麻雀算法/ISSA-XGBoostKey words
expressway/cost prediction/Lévy flight/improve the sparrow algorithm/ISSA-XGBoost分类
建筑与水利引用本文复制引用
李珏,刘洋..基于混合策略ISSA-XGBoost的高速公路工程造价预测研究[J].工程研究——跨学科视野中的工程,2026,18(1):70-84,15.基金项目
湖南省教育厅科研项目:基于"信息融合+价值链"的全过程工程咨询模式研究(22A0202) （22A0202）
