Abstract:
Gold is one of the most valuable metals in the world with numerous
applications, including jewelry and electronic devices. Countries often utilize gold as
an economic health indicator, and financial institutes use gold as a hedge against
loans. Additionally, gold is a popular investment asset for diversifying portfolios. Thus,
gold price predictions are crucial to making proper future decisions. Understanding the
reasons why the price of gold fluctuates is one of the challenging tasks. In earlier
research, gold price forecasting had been done using statistical methods. But with the
recent developments in machine learning methods, it is now possible to combine
conventional statistical models with machine learning to produce a hybrid model that
makes better predictions. In the first part of the work, a novel hybrid model was
proposed, by using Autoregressive Moving Average (ARIMA), Long-short Term
Memory (LSTM), and Prophet. It is also essential to develop models that can predict
gold prices during a crisis because, during these times, models will deviate from their
typical historical patterns. Hence, another attempt has been made to examine the
influence of crude oil, and silver, on gold prices during the 2008 financial crisis and the
COVID-19 period, and predict gold prices using regression Analysis, co-integration,
Vector Error Correction Model (VECM). It was found that there are short-term
causalities between gold and the previous month's crude oil and silver. Therefore,
having a joint impact on the current gold price during the crisis periods. Using the
models proposed in this paper, better gold price predictions can be made in the future,
even during financial crises. Better forecasting leads to better risk management,
investment decisions, hedging, economic analysis, and strategic trading giving the
opportunity to earn profits.
