Hybridization of hybrid-ARIMA-EM and XGBoost for enhanced price predictive modeling

Abstract

Managing finance entails the art and science of distributing available and potential funds among various competing needs. Government expenditures fund programs that provide a wide range of services to different population segments. As a result, the demand for enhanced and additional services often surpasses the government's financial capacity. Firstly, the price forecasting procedures for the extreme gradient boosting (XGBoost), gated recurrent unit (GRU), and hybrid-ARIMA-EM models will be summarized. Secondly, the accuracy of the models will be assessed on two real datasets collected from Kaggle (Crude_Oil_Price and KL_apartment). This study then proposes combining the hybrid-ARIMA-EM model with XGBoost to enhance the price forecasting performance in terms of time series analysis. Experimental results show that the suggested combination outperforms other selected models in price forecasting accuracy.