A gold price prediction model based on intelligent optimization and secondary decomposition techniques

As a representative of precious metals, gold has played an important role in finance, economy, culture, and other fields since ancient times. Accurate prediction of gold prices is critical to investment decisions, government regulation, and the operation of financial institutions. However, the complexity and non-stationarity of gold prices greatly hinder the further development of forecasting work. To solve the above problem, this paper innovatively proposes a quadratic decomposition forecasting model based on an optimization method, which aims to achieve deep mining and efficient forecasting of gold price time series data, so as to better cope with the uncertainty and complexity of the international market. The model initially combines crown-pig optimization (CPO), an intelligent swarm optimization algorithm, with an improved fully integrated Empirical Mode decomposition (iCEEMDAN) with adaptive noise to perform primary decomposition of the original signal. Then, the resulting highly complex subsequences are decomposed twice by variational mode decomposition (VMD) and then input into the BiLSTM network for accurate prediction. The experimental results show that this model has a better gold price prediction effect, with MAPE and R2 reaching 0.5082 and 0.9913, respectively, which is better than other comparison models. The hybrid prediction model proposed in this paper combines the CPO optimization algorithm with iCEEMDAN-VMD for gold price prediction, which provides a new way to improve prediction accuracy. This study will help governments, financial institutions, and investors to better understand the global economic situation, gain insight into potential market risks, and optimize their investment portfolio strategies accordingly.