Journal Search

Korean Journal of Construction Engineering and Management

Korean Journal of Construction Engineering and Management

ISO Journal Title : Korean J. Constr. Eng. Manag.
Open Access Journal Bimonthly
  • ISSN (Print) : 2005-6095
  • ISSN (Online) : 2465-9703
Online Submission

Current Issue : Vol.27 No.2, 2026-03

more

Development of a Prediction System for Construction Arbitral Awards Based on Cases Using Generative AI and BERT

Su Hyeon Choe ; Han Soo Kim

https://dx.doi.org/10.6106/KJCEM.2026.27.2.003

Construction projects are prone to disputes due to the involvement of multiple stakeholders and complex site conditions. Among these disputes, conflicts between clients and contractors occur most frequently, often leading to substantial financial losses. Arbitration provides a streamlined and legally binding resolution mechanism, and it is widely adopted in the construction sector. The objective of this study is to develop a system for predicting arbitral awards. To achieve this, the system incorporates GPT-based labeling to structure raw case data and BERTbased embeddings to automatically retrieve semantically similar arbitration cases, which then serve as the basis for prediction. A three-run repeated evaluation with identical inputs showed an average cosine similarity greater than 0.89 across outputs, thereby confirming the semantic consistency of the system’s predictions. These results demonstrate that the proposed approach can serve as a practical natural language-based tool for predicting arbitration award outcomes in the construction industry.

Full Text PDF Abstract

An Empirical Study on Factors Affecting Cost Overruns in BTO Road Projects

Ji-Eun Yang ; Jae-Sung Suk

https://dx.doi.org/10.6106/KJCEM.2026.27.2.016

Based on the Private Investment Promotion Act enacted in 1994, various public infrastructure projects in Korea have been implemented through private investment schemes. However, large-scale private investment projects are exposed to risks such as construction cost escalation and project delays due to their long implementation periods. These risks have weakened private sector participation, despite recent institutional reforms. Against this backdrop, this study empirically examines the factors influencing cost increases in BTO road projects, the most actively adopted form of private investment in the road sector. The potential presence of latent common factors among cost-related variables was first assessed using the Kaiser?Meyer?Olkin (KMO) measure and Bartlett’s test, followed by multiple regression analysis using the observed variables. The results provide empirical evidence to support policy design for revitalizing private investment projects and offer insights for developing project-specific risk management strategies.

Full Text PDF Abstract

Performance comparison of noise control algorithms for the point cloud-based 3D models

Dong-Gun Lee ; Han-Bin Park ; Kyu-Man Cho ; Tae-Hoon Kim

https://dx.doi.org/10.6106/KJCEM.2026.27.2.023

As smart construction technologies continue to expand in the construction industry, the utilization of point cloud data (PCD) has increased significantly. Consequently, preprocessing to remove noise inherently included during data acquisition has become essential. Among the commonly used noise-removal techniques, Statistical Outlier Removal (SOR) and Radius Outlier Removal (ROR) are widely adopted; however, the performance of these algorithms can vary substantially depending on the density of the acquired PCD. Therefore, this study quantitatively analyzes the influence of PCD density on the performance of SOR and ROR by comparing removal-rate sensitivity to parameter variations and the extent of normal-point loss. The results show that both algorithms exhibit relatively stable behavior in high-density PCD, with small variations in removal rate and minimal normal-point loss. In contrast, low-density PCD responds sensitively to parameter changes, leading to a significant increase in normalpoint loss. Furthermore, SOR, which is based on global distance statistics, demonstrated more stable performance than ROR in low-density conditions, whereas ROR showed slightly better stability in high-density environments due to its local neighbor-based decision process. These findings suggest that more cautious parameter tuning is required when applying SOR and ROR to low-density PCD. The outcomes of this study are expected to serve as fundamental reference data for selecting appropriate algorithms and determining parameter settings according to PCD density characteristics in future preprocessing workflows.

Full Text PDF Abstract

A Study on the Impact of International Exchange Complex Development on the Housing Market : Focusing on the Effect of Policy Environment on Auction Winning Bid Ratios

Min-Jeong Joo ; Sangyoub Lee

https://dx.doi.org/10.6106/KJCEM.2026.27.2.032

This study analyzes the multi-layered impacts of policy regulations?specifically financial restrictions and land transaction permit zones?and the development expectations of the International Exchange Complex (IEC) on apartment auction winning bid rates in Gangnam-gu. By integrating OLS, Machine Learning (ML), and Spatial Econometric (SAR) models, this research aimed to enhance both analytical robustness and predictive accuracy. The ML models (Random Forest, LightGBM) demonstrated superior predictive power (R²?0.507), validating the significance of complex, non-linear relationships within the market. Crucially, the SAR model effectively controlled for spatial autocorrelation found in OLS residuals. The non-significant spatial coefficient (ρ) supported the statistical stability of the OLS estimates, suggesting that spatial spillover effects are minimal in the auction market. Consequently, ... policy variables exhibited significant positive direct effects, implying that regulatory restrictions in the general market induced a balloon effect in the auction market. The results of this study first confirmed the potential for "regulation," a risk factor for development projects, to be transformed into "opportunities" depending on market conditions. Furthermore, it proposed criteria for defining the effective sphere of influence for large-scale urban development projects. Finally, this study, through an analytical approach combining machine learning and spatial econometric models, suggests the potential for enhancing a PropTech-based valuation model, offering important practical implications for the construction and real estate development sectors.

Full Text PDF Abstract

Performance Analysis of Corner Extraction Methods for the 3D Point Cloud Building Models - Focusing on RANSAC and Convex Hull Approaches -

Han-Bin Park ; Dong-Gun Lee ; Kyu-Man Cho ; Tae-Hoon Kim

https://dx.doi.org/10.6106/KJCEM.2026.27.2.042

With the increasing use of point cloud data (PCD) for building facade inspection and digital-twin construction, the importance of reliable corner detection techniques has continued to grow. However, the density of PCD acquired in field environments varies significantly depending on sensor performance, scanning conditions, and accessibility, and the impact of such density variations on corner detection accuracy has not been sufficiently verified. To address this issue, this study constructed high-density and low-density PCD for the same building and quantitatively compared the corner detection performance of two representative point-extraction algorithms: RANSAC and Convex Hull. The results show that both methods yielded stable corner positions in high-density PCD, whereas low-density PCD exhibited accuracy degradation due to geometric distortion and point omission. These findings demonstrate that PCD density is a critical factor influencing the reliability of corner detection and highlight the need for further research to improve feature extraction robustness under low-density scanning conditions.

Full Text PDF Abstract

Simulation-based Evacuation Risk Assessment Framework of Deep Subway Station ? Case Study of Guryong Station

Ji-Won Jeong ; Fansheng Kong ; Seung-Jun Ahn

https://dx.doi.org/10.6106/KJCEM.2026.27.2.051

This study proposes a generalized analysis framework to quantitatively evaluate latent evacuation vulnerabilities within the complex vertical egress system of deep-depth subway stations by adapting the International Maritime Organization (IMO) passenger ship evacuation analysis guidelines (IMO MSC/Circ.1238). Departing from conventional evaluations centered on single macroscopic indicators, this research establishes a probabilistic scenario-setting procedure and introduces the Bottleneck Sensitivity Index (BSI) as a key metric to quantify the bottleneck levels experienced by users based on their initial departure locations. A case study conducted on Guryong Station (Suin-Bundang Line) demonstrated that the proposed framework can specifically identify localized vulnerability points for users on lower floors when vertical path availability changes due to obstructions on upper levels. By successfully adapting recognized maritime guidelines for land transportation facilities, this study lays the foundation for objective and consistent safety diagnostic standards. Furthermore, the standardized procedures and intuitive analysis methods ensure high practical utility for designers and regulators, making the framework a vital decision-support tool for future safety design and the development of efficient evacuation guidance strategies in deep-depth subway stations.

Full Text PDF Abstract

Development of Deep Learning Methodology of Object and Action Recognition for Discrete Event Simulation in Construction

Hyeonguk Choi ; Yongho Ko ; Gihun Kim ; Seungwoo Han

https://dx.doi.org/10.6106/KJCEM.2026.27.2.061

Enhancing the productivity and detecting inefficiency of earthmoving equipment are key factors leading to successful project delivery. This requires sufficient time and effort of data collection and simulation to optimize equipment fleet. This study proposes a deep learning?based framework that combines video-based action recognition with Web CYCLONE simulation for automatic activity time collection that is used as inputs of discrete event simulation model of earthmoving operation. CCTV footage collected from an on-going road construction site was collected. Excavators were detected, tracked, and classified into four actions?excavating, rotating, loading, and returning?through an integrated Faster R-CNN, SORT, and BiLRCN approach. The extracted action durations were applied to Web CYCLONE to estimate cycle time and productivity. Simulation results showed an average cycle of about 175 seconds, with the hauling phase occupying 85% of the total duration, confirming it as the dominant productivity factor. The proposed method demonstrates that automated video analysis can identify process bottlenecks and support data-driven productivity assessment in smart construction.

Full Text PDF Abstract