This paper aims to use the tree-based methods for time series data forecasting and compare between Decision Tree (DT), Random Forest (RF), Gradient Boosted Trees (GBT) and ARIMA model to predict monthly gold prices. The time series data for the monthly gold prices was used during the period from Nov-1989 to Dec-2019, which represents 362 observations. ARIMA, DT, RF, and GBT models were fitted based on 90% of data as training set. Then, their accuracy was compared using the statistical measure RMSE. The results indicated that RF was better than DT, GBT and ARIMA (0,1,1) in predicting future gold prices, based on RMSE= 38.52.
Digital Object Identifier (DOI)
Houssainy A. Rady, EL; Fawzy, Haitham; and Mohamed Abdel Fattah, Amal
"Time Series Forecasting Using Tree Based Methods,"
Journal of Statistics Applications & Probability: Vol. 10
, Article 21.
Available at: https://dc.naturalspublishing.com/jsap/vol10/iss1/21