Comparison of commodity prices by using machine learning models in the COVID-19 era

Abstract

Commodity products such as gold, silver, and metal have been seen as safe havens in past economic crises. This situation increases the interest in commodity products. Due to the COVID-19 pandemic, quarantine decisions and precautions have caused an economic slowdown in stock markets and consumer activities. This inactivity in the economy has led to the COVID-19 recession that started in February 2020. Because of the increase in the number of COVID-19 cases, the difficulty of physical buying-selling transactions has shown that commodity products can be a safe investment tool. Based on the fact that machine learning approaches gained importance in commodity price prediction, the main goal of this study is to understand whether machine learning methods are meaningful for commodity price prediction even in extraordinary situations. To measure commodities’ price volatility, a data set obtained from Borsa İstanbul is separated into pre-COVID-19 and COVID-19 periods. Daily prices for gold and silver commodities, from July 2018, which is before the ongoing COVID-19 recession, to October 2021 are used. The performances of the machine learning models were compared with MAE, MAPE, and RMSE metrics. The findings of this study point out that the LSTM model has more accurate predictions, especially in the pre-COVID-19 period. When considering the COVID-19 period only, SVR produces the best prediction results for the gold commodity and LSTM has the best prediction results for the silver commodity.