Performance Analysis of Standard Construction Code Matching Model Using Ensemble Techniques

This study aims to address the inaccuracies in cost estimation and increased project management complexity arising from the inefficient use of standardized construction codes in the construction industry. To achieve this, a machine learning-based model was developed to improve the accuracy of automatic matching between construction specifications and standard codes. Specifically, this study evaluated the effectiveness of three ensemble techniques—stacking, bagging, and boosting—across five levels of data complexity, reflecting the hierarchical structure and varying complexity of standard construction codes. Results indicated that all ensemble models outperformed the base model, with bagging and boosting showing high accuracy in higher- level codes (Levels 1-3). However, performance improvements were limited for lower-level codes (Levels 4-5). These findings suggest that selecting techniques tailored to data complexity can optimize matching accuracy and enhance the potential for automation in construction code applications. Ultimately, precise matching of standard codes to construction items is expected to improve the efficiency and accuracy of construction cost management.