School of Economics and Management, China University of Geosciences, Wuhan, Hubei, China.
Economics & Technology Research Institute, China National Petroleum Corporation, Beijing, China.
PLoS One. 2023 Oct 30;18(10):e0285631. doi: 10.1371/journal.pone.0285631. eCollection 2023.
Copper is an important mineral and fluctuations in copper prices can affect the stable functioning of some countries' economies. Policy makers, futures traders and individual investors are very concerned about copper prices. In a recent paper, we use an artificial intelligence model long short-term memory (LSTM) to predict copper prices. To improve the efficiency of long short-term memory (LSTM) model, we introduced a simulated annealing (SA) algorithm to find the best combination of hyperparameters. The feature engineering problem of the AI model is then solved by correlation analysis. Three economic indicators, West Texas Intermediate Oil Price, Gold Price and Silver Price, which are highly correlated with copper prices, were selected as inputs to be used in the training and forecasting model. Three different copper price time periods, namely 485, 363 and 242 days, were chosen for the model forecasts. The forecast errors are 0.00195, 0.0019 and 0.00097, respectively. Compared with the existing literature, the prediction results of this paper are more accurate and less error. The research in this paper provides a reliable reference for analyzing future copper price changes.
铜是一种重要的矿物质，其价格波动会影响一些国家经济的稳定运行。政策制定者、期货交易员和个人投资者都非常关注铜价。在最近的一篇论文中，我们使用人工智能模型长短期记忆（LSTM）来预测铜价。为了提高长短期记忆（LSTM）模型的效率，我们引入了模拟退火（SA）算法来寻找最佳的超参数组合。然后通过相关性分析来解决 AI 模型的特征工程问题。选择了与铜价高度相关的三个经济指标，即西德克萨斯中质原油价格、黄金价格和白银价格，作为输入，用于训练和预测模型。模型对三个不同的铜价时间段（485、363 和 242 天）进行了预测。预测误差分别为 0.00195、0.0019 和 0.00097。与现有文献相比，本文的预测结果更加准确，误差更小。本文的研究为分析未来铜价变化提供了可靠的参考。
