基于TTHHO-LSSVM-ABKDE的装配式混凝土建筑造价预测方法

Research on cost prediction of prefabricated concrete buildings based on TTHHO-LSSVM-ABKDE

摘要

针对装配式混凝土建筑造价预测存在的非线性、小样本难题,以及传统模型精度不足、泛化能力弱的局限,先对标准哈里斯鹰优化算法(HHO)进行改进,再引入 ABKDE 对 LSSVM 预测残差的非对称、异方差分布进行动态建模,构建出融合改进瞬态三角哈里斯鹰优化算法(TTHHO)、最小二乘支持向量机(LSSVM)与自适应带宽核密度估计(ABKDE)的TTHHO-LSSVM-ABKDE 预测模型;然后基于装配式混凝土建筑项目数据开展实证分析,通过消融实验、五折交叉验证及与 PSO-LSSVM、NGO-LSSVM、CPO-LSSVM 等模型的对比验证,结果表明:所提模型测试集决定系数R²达0.986 43,平均绝对百分比误差(MAPE)降至6.14%,五项核心误差指标(RMSE、MAE、MBE、MAPE、MSE)较原始 LSSVM 模型降幅达 92.96%~95.02%;五折交叉验证平均R²稳定在 0.954 5,收敛速度较对比模型提升显著。该模型有效克服传统预测方法的局限性,为装配式混凝土建筑造价的精准预测提供了新工具,同时丰富了智能优化算法与机器学习模型融合的理论体系,对推动装配式建筑规模化高质量发展具有重要实践价值。

Abstract

To address the nonlinearity and small-sample challenges in cost prediction of prefabricated concrete buildings, as well as the limitations of traditional models in terms of insufficient accuracy and weak generalization ability, this study first improves the standard Harris Hawks Optimization (HHO) algorithm, and then introduces Adaptive Bandwidth Kernel Density Estimation(ABKDE) to dynamically model the asymmetric and heteroscedastic distribution of prediction residuals from Least Squares Support Vector Machine (LSSVM). A hybrid prediction model, designated as TTHHO-LSSVM-ABKDE, is thereby constructed by integrating the improved Transient Trigonometric Harris Hawks Optimization (TTHHO) algorithm, LSSVM, and ABKDE. Empirical analysis is carried out based on the data of prefabricated concretebuildings, and through ablation experiments, 5-fold cross-validation, and comparative validation against models such as PSO-LSSVM, NGO-LSSVM, and CPO-LSSVM, the results show that the proposed model achieves a coefficient of determination coefficient R² of 0.986 43 on the test set, with the mean absolute percentage error (MAPE) reduced to 6.14%. The five core error indicators (RMSE, MAE, MBE, MAPE, MSE) are reduced by 92.96% to 95.02% compared with the original LSSVM model. The average R² from 5-fold cross-validation remains stable at 0.954 5, and the convergence speed is significantly improved over that of the comparative models. This model effectively overcomes the limitations of traditional prediction methods, provides a new tool for accurate prediction of the cost of prefabricated concrete buildings, and enriches the theoretical framework for integrating intelligent optimization algorithms with machine learning models, thereby holding important practical value for promoting the large-scale and high-quality development of prefabricated buildings.

关键词

改进的是瞬态三角哈里斯鹰 / 最小二乘支持向量机 / 自适应带宽核密度估计 / 造价预测

Key words

improved transient triangular Harris hawk optimization algorithm / least squares support vector machine / adaptive bandwidth kernel density estimation / cost prediction

[1] 唐碧秋,许仕铭,邹洋,等.基于GCA-TOPSIS的装配式建筑建造成本影响因素研究[J].混凝土,2024(5):135-141.
Tang Biqiu, Xu Shiming, Zou Yang, et al. Research on influencing factors of construction cost of prefabricated buildings based on GCA-TOPSIS[J]. Concrete, 2024(5):135-141.
[2] 刘瑶,宋红,徐沛.基于AHP-熵权法的装配式建筑成本影响因素研究[J].建筑经济,2023,44(S1):150-154.
Liu Yao, Song Hong, Xu Pei. Research on influencing factors of prefabricated building cost based on AHP-entropy weight method[J]. Construction Economics, 2023,44(S1):150-154.
[3] 罗祥,黄喜兵.基于ISM的装配式建筑成本控制研究[J].建筑经济,2021,42(11):73-77.
Luo Xiang, Huang Xibing. Research on cost control of prefabricated buildings based on ISM[J]. Construction Economics, 2021,42(11):73-77.
[4] 杨柳荣.基于BIM技术的装配式建筑全过程造价控制研究[J].工程抗震与加固改造,2025,47(2):203.
Yang Liurong. Research on whole-process cost control of prefabricated buildings based on BIM technology[J]. Earthquake Resistant Engineering and Retrofitting, 2025,47(2):203.
[5] 申琪玉,李杰杰.基于不同装配率的装配式混凝土建筑投资估算指标差异与影响因素研究[J].建筑经济,2021,42(S1):181-185.
Shen Qiyu, Li Jiejie. Research on differences and influencing factors of investment estimation indicators for prefabricated concrete buildings with different assembly rates[J]. Construction Economics, 2021,42(S1):181-185.
[6] 林申正,罗恒勇.BIM技术和神经网络相融合的装配式建筑成本控制研究[J].建筑经济,2021,42(2):86-89.
Lin Shenzheng, Luo Hengyong. Research on cost control of prefabricated buildings integrating BIM technology and neural network[J]. Construction Economics, 2021,42(2):86-89.
[7] 李丽红,李广新,雷云霞,等.人材机价格波动对装配式建筑构件造价的影响研究——以沈阳市为例[J].建筑经济,2019,40(8):81-84.
Li Lihong, Li Guangxin, Lei Yunxia, et al. Research on the impact of price fluctuations of labor, materials and machinery on the cost of prefabricated building components——Taking Shenyang as an example[J]. Construction Economics, 2019,40(8):81-84.
[8] 王林峰,刘云,亓彦珣,等.基于IPSO-LSSVR算法的变电站工程造价预测方法[J].沈阳工业大学学报,2025,47(2):168-175.
Wang Linfeng, Liu Yun, Qi Yanjun, et al. Substation project cost prediction method based on IPSO-LSSVR algorithm[J]. Journal of Shenyang University of Technology, 2025,47(2):168-175.
[9] 金季岚.基于马尔科夫链的建筑工程造价预测模型[J].兵工自动化,2024,43(9):32-35.
Jin Jilan. Construction project cost prediction model based on Markov chain[J]. Ordnance Industry Automation, 2024,43(9):32-35.
[10] 邓宇轩.基于自适应PSO-BP神经网络的城轨造价预测研究[J].建筑经济,2023,44(S1):92-96.
Deng Yuxuan. Research on urban rail cost prediction based on adaptive PSO-BP neural network[J]. Construction Economics, 2023,44(S1):92-96.
[11] 谢金豪,刘文昌.基于GA-BP神经网络的建筑工程造价预测研究[J].建筑经济,2022,43(S1):235-240.
Xie Jinhao, Liu Wenchang. Research on construction project cost prediction based on GA-BP neural network[J]. Construction Economics, 2022,43(S1):235-240.
[12] 鲁淑霞,田如娜.结构化加权最小二乘支持向量机[J].计算机科学,2013,40(12):52-54.
Lu Shuxia, Tian Runa. Structured weighted Least Squares Support Vector Machine[J]. Computer Science, 2013,40(12):52-54.
[13] Zougab N,Adjabi S,Kokonendji C C.Bayesian estimation of adaptive bandwidth matrices in multivariate kernel density estimation[J].Computational Statistics & Data Analysis,2014,75:28-38.
[14] Wandmp,Jonesmc.Kernelsmoothing[M].Boca Rtona:Chemical Rubber Company Press,1994.
[15] 贾科,杨哲,魏超,等.基于斯皮尔曼等级相关系数的新能源送出线路纵联保护[J].电力系统自动化,2020,44(15):103-111.
Jia Ke, Yang Zhe, Wei Chao, et al. Pilot protection for renewable energy transmission lines based on Spearman rank correlation coefficient[J]. Automation of Electric Power Systems, 2020,44(15):103-111.
[16] Abdulrab H,Hussin F A,Ismail I,et al.Energy efficient optimal deployment of industrial wireless mesh networks using transient trigonometric Harris Hawks optimizer[J].Heliyon,2024,10(7):e28719.
[17] 李寿德.一种新的单纯形法—最大最小法[J].青海师专学报,1992(4):54-58.
Li Shoude. A new simplex method—Max-min method[J]. Journal of Qinghai Junior Teachers‘ College, 1992(4):54-58.
[18] 胡山,朱向华.基于ResNet18与K折交叉验证的电机轴承故障诊断[J].自动化与仪表,2025,40(4):40-45.
Hu Shan, Zhu Xianghua. Motor bearing fault diagnosis based on ResNet18 and K-fold cross-validation[J]. Automation & Instrumentation, 2025,40(4):40-45.
[19] 刘春茂,郝倩,张云岗.基于PSO-LSSVM的网络流量预测[J].微型电脑应用,2016,32(5):27-30.
Liu Chunmao, Hao Qian, Zhang Yungang. Network traffic prediction based on PSO-LSSVM[J]. Microcomputer Applications, 2016,32(5):27-30.
[20] 王璜瑞,汪伟,祁云,等.基于CPO-LSSVM的煤层底板突水量等级预测模型[J].沈阳理工大学学报,2025,44(5):82-88.
Wang Huangrui, Wang Wei, Qi Yun, et al. Prediction model of coal seam floor water inrush volume grade based on CPO-LSSVM[J]. Journal of Shenyang Ligong University, 2025,44(5):82-88.
[21] 马亮.改进北方苍鹰优化LSSVM的高程异常拟合算法[J].地理空间信息,2025,23(3):18-21.
Ma Liang. Elevation anomaly fitting algorithm using improved northern goshawk optimization for LSSVM[J]. Geospatial Information, 2025,23(3):18-21.

基金资助

新疆生产建设兵团重点领域科技攻关计划项目(2021AB027)