eISSN:2148-0710 - pISSN:1301-6229
Home  |  Legal   |  News  |  Contact   |  Türkçe

Quick access


Prediction of Gold Prices by Xgboost and Mars Methods
As gold is an important means of payment, investment and savings, determination of prices is important for countries and investors. Therefore, the prediction of the gold price is aimed in this study. For this purpose, the variables such as silver price, crude oil WTI futures price, US Dollar index, S&P500 index, US federal funds compound interest rate, and US CPI which are thought to have effect on the gold price, were used as inputs in the models. The data used belongs to the period January 2015 - June 2020. Gold price is a non-linear series, besides it is non-stationary. These features of the gold price make it difficult to obtain price predictions. For this reason, it is appropriate to use machine learning methods and non-parametric methods in prediction of the gold price in addition to classical methods. In this study, XGBoost, MARS and linear regression models were used to obtain the predictions. The results obtained were compared using the performance evaluation criteria of the models, and the effects of the input variables on the gold price for the XGBoost and MARS models were determined. Among the models used, the XGBoost model provided the most successful results with a 99.6% successful prediction rate. For the MARS model, this rate is 97.8%. These ratios show that the variables used have a significant effect on gold prices. Among the variables used, the variable that has the most important effect on gold prices is the US CPI. In addition, the findings show that the XGBoost and MARS methods are preferable methods to obtain estimates for gold price and similar series.

Gold Price, Prediction, Machine Learning, Nonparametric Regression, XGBoost, MARS.

Advanced Search


Adress :
Phone : Fax :
Email :dergi@ekevakademi.org

Web Yazılım & Programlama Han Yazılım Bilişim Hizmetleri