Time series forecasting using gradient boosting algorithms

Abstract

This study investigates the efficiency of gradient boosting algorithms, particularly XGBoost, in time series forecasting. We optimize the parameters using RandomizedSearchCV and apply the model to daily stock prices of the Ethereum cryptocurrency. Additionally, we compare the prediction performance of XGBoost with two other models, LightGBM and CatBoost. Our findings reveal that the LightGBM model outperforms both CatBoost and XGBoost in terms of accuracy for time series prediction