| Title |
Resolving Multicollinearity Issues in Construction Cost Driver Analysis With Limited Data |
| Authors |
김진희(Kim, Jin-Hee,) ; 김재식(Kim, Jae-Sik) ; 허영기(Huh, Young-Ki) |
| DOI |
https://doi.org/10.5659/JAIK.2024.40.10.269 |
| Keywords |
Construction Cost of Retail Building; Factors; Multicollinearity; VIF; PCA; LASSO; Ridge Regression Analysis |
| Abstract |
Factors influencing the construction costs of large-scale retail buildings in Korea were analyzed to address multicollinearity issues using
limited data. Various statistical techniques, including VIF, PCA, LASSO, and Ridge regression, were applied and compared to improve the
reliability of the analysis. Key variables such as the number of parking spaces, building area, and the number of above-ground floors were
identified as major factors affecting total construction costs. PCA was used to transform the main components, effectively reducing
multicollinearity and providing stable results. Bootstrapping and cross-validation methods were employed to assess the robustness of the
models. Ridge regression outperformed LASSO regression in delivering reliable insights under limited data conditions. |