Gold Futures Price Prediction Using Transformer Deep Learning Models with Data Scraped via UiPath

Area of Science:

• Economics
• Financial Markets
• Data Science

Background:

• Gold is a globally traded commodity crucial for economic stability and central bank reserves.
• Predicting gold prices is complex due to market volatility and numerous influencing factors.
• Accurate gold price forecasting is vital for investment strategies and economic policy.

Purpose of the Study:

• To develop and evaluate a transformer model for predicting daily gold prices.
• To assess the model's accuracy and efficiency compared to traditional forecasting methods.
• To provide a reliable tool for decision-making in the gold bullion market.

Main Methods:

• Web scraping of daily gold prices from investing.com using UiPath (Robotic Process Automation).
• Data preprocessing including missing value imputation and MinMax scaling.
• Implementation and training of a transformer model for time-series price prediction.

Main Results:

• The transformer model achieved high prediction accuracy with a Mean Squared Error (MSE) of 0.0224 and R-squared of 0.93.
• The model demonstrated proficiency in detecting both short-term price fluctuations and long-term market trends.
• Performance metrics indicate superior accuracy and dependability over conventional forecasting techniques.

Conclusions:

• The transformer model offers a more accurate and dependable method for gold price forecasting.
• The study provides valuable guidance for investors, financial analysts, and policymakers.
• Future enhancements could incorporate sentiment analysis from news and social media for richer insights.