Research on Price Prediction of Construction Materials Based on VMD-SSA-LSTM

Abstract

This study proposes a novel hybrid model, VMD-SSA-LSTM, aimed at enhancing the accuracy of construction material price (CMP) predictions. The model integrates Variational Mode Decomposition (VMD) for signal decomposition, the Sparrow Search Algorithm (SSA) for parameter optimization, and Long Short-Term Memory (LSTM) networks for predictive modeling. Historical CMP data are first decomposed into intrinsic components using VMD, followed by the SSA-based optimization of the LSTM parameters. These components are then input into the LSTM network for final predictions, which are aggregated to produce the CMP forecast. Experimental results using rebar price data from Hengyang City demonstrate that the VMD-SSA-LSTM model outperforms the backpropagation (BP) neural network, LSTM, and VMD-LSTM models in terms of prediction accuracy. The proposed method provides highly valuable tools for construction cost management, significantly enhancing the reliability of budget planning and risk mitigation decisions, and has significant practical implications for engineering cost risk management.