Copper price prediction using LSTM recurrent neural network integrated simulated annealing algorithm.

Area of Science:

• Economics
• Artificial Intelligence
• Time Series Analysis

Background:

• Copper price fluctuations significantly impact national economies, concerning policymakers, traders, and investors.
• Accurate copper price prediction is crucial for economic stability and investment strategies.
• Existing prediction models often lack the precision required for dynamic market conditions.

Purpose of the Study:

• To develop a highly accurate AI model for predicting copper prices.
• To enhance the efficiency of the long short-term memory (LSTM) model through hyperparameter optimization.
• To provide a reliable tool for analyzing future copper price trends.

Main Methods:

• Utilized an artificial intelligence approach, specifically the long short-term memory (LSTM) network.
• Integrated a simulated annealing (SA) algorithm to optimize LSTM hyperparameters for improved performance.
• Employed correlation analysis for feature engineering, selecting highly correlated economic indicators (WTI Oil, Gold, Silver prices) as model inputs.
• Trained and forecasted copper prices across three distinct time periods (485, 363, and 242 days).

Main Results:

• Achieved highly accurate copper price forecasts with minimal errors: 0.00195 (485 days), 0.0019 (363 days), and 0.00097 (242 days).
• Demonstrated superior prediction accuracy compared to existing literature.
• Validated the effectiveness of the combined LSTM and SA approach for time series forecasting.

Conclusions:

• The developed AI model offers a reliable and accurate method for copper price prediction.
• The integration of simulated annealing and correlation analysis significantly enhances LSTM model performance.
• This research provides valuable insights for stakeholders involved in copper markets and economic policy.