Scholarlink

<p>Construction duration prediction is crucial for the early-stage planning of ultra-high-rise building (UHRB) projects, but its accuracy is often limited by high-dimensional, strongly correlated, and mixed-type engineering features. To overcome these issues, this study develops a hybrid PCA-ISSA-BPNN framework that combines principal component analysis (PCA), an improved sparrow search algorithm enhanced with grey wolf optimization (ISSA), and a back-propagation neural network (BPNN). PCA is used to reduce multicollinearity and redundancy in the quantitative variables, while ISSA performs global hyperparameter optimization for the BPNN. Based on 87 UHRB projects and five-fold cross-validation, the proposed model achieved the best predictive performance among all compared methods, with <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{R}^{2}=0.982\)</EquationSource> </InlineEquation>, RMSE = 23.496, MAE = 15.742, and MAPE = 2.0%. These results demonstrate that the proposed framework can simultaneously improve accuracy, robustness, and feature representation quality. The PCA-ISSA-BPNN model therefore provides a practical and effective tool for early-stage duration estimation and schedule-risk screening in UHRB projects.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Construction duration prediction for ultra-high-rise buildings using PCA and an improved sparrow-search–optimized BPNN

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月14日

查看全文
  • Fei-yang Yuan,
  • Wen-tong Lyu,
  • Pei-juan Zheng,
  • Hou-jun Li,
  • Yong-qi Liu,
  • Rony Badhon Roy,
  • Xian-chao Yang,
  • Xu Feng

摘要

Construction duration prediction is crucial for the early-stage planning of ultra-high-rise building (UHRB) projects, but its accuracy is often limited by high-dimensional, strongly correlated, and mixed-type engineering features. To overcome these issues, this study develops a hybrid PCA-ISSA-BPNN framework that combines principal component analysis (PCA), an improved sparrow search algorithm enhanced with grey wolf optimization (ISSA), and a back-propagation neural network (BPNN). PCA is used to reduce multicollinearity and redundancy in the quantitative variables, while ISSA performs global hyperparameter optimization for the BPNN. Based on 87 UHRB projects and five-fold cross-validation, the proposed model achieved the best predictive performance among all compared methods, with \(\:{R}^{2}=0.982\) , RMSE = 23.496, MAE = 15.742, and MAPE = 2.0%. These results demonstrate that the proposed framework can simultaneously improve accuracy, robustness, and feature representation quality. The PCA-ISSA-BPNN model therefore provides a practical and effective tool for early-stage duration estimation and schedule-risk screening in UHRB projects.