Predicting construction cost index using a VMD-GRU framework with the mhsa mechanism

The Construction Cost Index (CCI), published monthly by Engineering News-Record in the US, has fluctuations that create significant challenges in cost estimation, bid preparation, and investment planning. Accurate forecasting of the CCI is critical for improving the precision of cost estimates and capital project budgets, thereby facilitating the preparation of more reliable and competitive bids. To address these challenges, a hybrid forecasting model is introduced in this study, integrating variational mode decomposition (VMD), gated recurrent unit (GRU), and multihead self-attention (MHSA) mechanisms to predict the CCI over various time horizons. The VMD technique decomposes the CCI time series into intrinsic mode functions (IMFs), effectively capturing distinct frequency components and mitigating the nonstationary characteristics of the data. The GRU captures temporal dependencies and extracts essential features. The MHSA mechanism enhances the model's capability to focus on different segments of the input sequence, improving feature extraction and preserving contextual information. The effectiveness of the proposed model was assessed using CCI data from California and benchmarked against several conventional forecasting models. The results consistently demonstrate that the hybrid VMD-GRU-MHSA model achieves superior prediction accuracy across short-term, medium-term, and long-term forecasts. The application of hybrid models in construction cost time-series forecasting is advanced, and a novel methodology has been introduced for enhancing budget estimation and financial planning in the construction sector. By improving prediction accuracy, the VMD-GRU-MHSA framework strengthens financial management, supports informed decision-making, and drives strategic planning efforts.