The purpose of this study is to enhance the accuracy and stability of predictions for the Architecture Billings Index (ABI), a critical economic indicator in the US construction industry. Existing forecasting methods, including traditional time-series and statistical learning models, often fail to fully capture the ABI's highly volatile characteristics, leading to limited accuracy and robustness. By addressing the challenges posed by the nonlinearity, non-stationarity, and noise characteristics of ABI data, the study aims to improve forecasting performance and offer a robust solution for both short-term and long-term predictions.
The integrated framework combines the Variational Mode Decomposition (VMD), Bidirectional Gated Recurrent Unit (BiGRU), and Grey Wolf Optimizer (GWO) techniques. First, the VMD method is applied to decompose the original ABI sequence into multiple-scale Intrinsic Mode Functions (IMFs); then, a BiGRU model is trained on each IMF to capture the complex nonlinear patterns and dependencies in the time series. Finally, the GWO algorithm is used to globally optimize the key hyperparameters of the BiGRU model. The validation of the predictive hybrid model involved three quantitative metrics of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE).
The empirical results show that the proposed model significantly outperforms traditional models in both short-term and long-term forecasting of the ABI. In short-term forecasts, it reduces RMSE by approximately 63% compared to the BiGRU model (1.2542 vs. 3.4445). In long-term forecasting, the proposed model reduces RMSE by about 34% compared to the BiGRU model (3.3905 vs. 5.1311).
This study introduces a novel VMD-GWO-BiGRU prediction framework that advances the theoretical and practical understanding of forecasting the ABI. By integrating VMD, GWO, and BiGRU, the framework leverages the strengths of each component to overcome the limitations of existing methods in accurately capturing the ABI, providing a more robust forecasting solution and addressing a critical methodological gap. The framework's validation demonstrates its practical applicability and potential in assisting practitioners monitor market dynamics and anticipate future cost for architectural services and construction activities.
