Title Construction Cost Index Forecasting Through Multivariate Time Series Modeling and Leading Indicators
Authors 윤하현(Yoon, Ha-Hyeon) ; 허찬(Heo, Chan) ; 박문서(Park, Moon-Seo) ; 안창범(Ahn, Changbum)
DOI https://doi.org/10.5659/JAIK.2025.41.5.319
Page pp.319-329
ISSN 2733-6247
Keywords Construction Cost Index; Machine Learning; Multivariate Model; Statistical Validation; Leading Indicators
Abstract The Construction Cost Index (CCI) is a key measure of price fluctuations in major construction resources, playing a crucial role in cost estimation and price trend analysis. Accurate CCI forecasting is essential to prevent cost underestimation or overestimation, ensuring the economic feasibility of construction projects. This study forecasts the CCI using a multivariate time series model, Vector Autoregression (VAR), to address the limitations of univariate models, especially during economic uncertainty. Through statistical validation, three leading indicators were identified: construction order amount, business survey index (BSI), and producer price index (PPI) for structural steel. The proposed model was optimized using the Akaike Information Criterion (AIC), while benchmark models ARIMA, VAR (CPI, PPI), and SVR were optimized through grid search. Model validation was conducted using data from January 2000 to April 2023, segmented into three economic phases: stability, heightened uncertainty, and a combined period. Walk-forward cross-validation assessed predictive performance over short-term forecasts of 3 months, mid-term forecasts of 6 months, and long-term forecasts of 12 months, with evaluation based on averaged performance metrics over multiple iterations. Results showed that the proposed model achieved the lowest error and highest accuracy in shortand mid-term forecasts. For long-term forecasts, SVR recorded the lowest error; however, qualitative analysis indicated that the proposed model more effectively captured overall trends in a balanced manner. By integrating key market indicators, this approach provides a robust method for CCI forecasting, enhancing cost predictability in the construction industry.