Prediction of construction material prices based on the sliding window catboost model

Material prices significantly impact project costs, and accurate price forecasting is crucial for developing cost management strategies. The volatility of construction material prices exhibits nonlinear and dynamic characteristics, posing challenges for prediction. This study proposes a price forecasting model for threaded steel based on the sliding window technique and the CatBoost algorithm. The proposed model was compared with the original CatBoost, the sliding window eXtreme Gradient Boosting (the SW XGBoost), the Auto-Regressive Integrated Moving Average (ARIMA), the Attention-based Convolutional Neural Network-Long Short-Term Memory (ATT-CNN-LSTM), and the nonlinear neural network models. The results demonstrated that the sliding window-based CatBoost model achieves excellent prediction accuracy, particularly during the 2020 pandemic, where the predicted curve closely matches the actual data. This study contributes to the current body of knowledge by presenting a novel approach that integrates the sliding window technique with the CatBoost algorithm to effectively capture the nonlinear and dynamic nature of construction material price fluctuations. The findings of this study are beneficial to practice as well, because the improved accuracy and reliability of price predictions can assist construction professionals in making informed decisions regarding cost management, budgeting, and procurement strategies.