Year: 2026
Type: article
Abstract: There are a wide variety of investment instruments in the time we live in. Among these instruments, gold is the oldest investment instrument. The Dollar index is an index that shows the strength of the dollar against a major basket of the most popular currencies. The Nasdaq 100 Index is an index of the world's largest non-financial companies listed on the Nasdaq stock exchange. Bitcoin was the first cryptocurrency to emerge in 2009 and has been a popular investment vehicle in recent years. In this study, both the default and hyperparameter-optimized versions of five machine learning techniques — Linear Regression, Lasso, Decision Tree, Random Forest, and XGBoost — were employed to forecast the values of Gold, Bitcoin, the U.S. Dollar Index, and the Nasdaq 100 Index over two-, five-, and ten-year periods spanning 2015–2025. One currency is selected as the dependent variable, and three currencies are selected as independent variables. The dependent variable (currency) is estimated using the independent variables. The predictive capabilities of the models were evaluated using the RMSE, MSE, MAE, and R² performance metrics. According to the results, the Opt_Lasso model achieved the best forecasting performance among all models. A review of the existing literature reveals that most prior studies have primarily focused on predicting just a single dependent variable using various independent variables. In contrast, the present study distinguishes itself by performing simultaneous forecasts for four different currencies within a single application framework, thereby extending beyond previous research and offering a novel contribution to the field of multivariate prediction modeling.
Authors: Hazim İşcan, Samet Kaan Kanak
Institution: Konya Technical University
Language: Turkish
Field: Decision Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7131353127