Advanced Techniques for Daily Gold Price Forecasting in India Through Statistical Analysis and Predictive Modeling

Gold, a crucial asset in the Indian economy, is influenced by numerous factors, including economic indicators and market volatility. The methodology adopted in this study encompasses a comprehensive preprocessing framework tailored for the TFT model, effectively addressing challenges such as missing timesteps and incorporating key variables like the Consumer Price Index (CPI), USD to INR exchange rate, and other macroeconomic indicators that influence gold prices. The study involved conducting a thorough statistical analysis to understand the impact of these variables on gold prices, enabling the careful selection of features suitable for the model. A meticulous tuning of model parameters, including the learning rate, dropout rates, and hidden size, was carried out through a series of experimental trials aimed at optimizing performance. Each trial was evaluated based on several metrics, including Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE), leading to significant improvements in the model's predictive capabilities. Additionally, various baseline models such as ARIMA, Linear Regression, and LSTM were benchmarked against the TFT model using a dataset spanning from 2011 to September 2024. The findings reveal that the TFT model not only surpasses traditional forecasting methods but also provides an efficient framework for understanding the impact of economic fluctuations on gold prices. The study also emphasizes the importance of interpretability in forecasting models, utilizing techniques like attention mechanisms and variable importance analysis to offer deeper insights into the model's decision-making process.