An Application of AdaBoost-GRU Ensemble Model to Economic Time Series Prediction

Abstract

Objectives: Given the importance of accurate prediction of financial time seriesdata and their benefits in the real-life, AdaBoost-GRU ensemble learning isproposed in which it’s forecasting accuracy is to be compared with AdaBoostLSTM, single Long Short Term Memory (LSTM), and single Gated RecurrentUnit (GRU). Methods: The data for Korea Composite Stock Price Index (KOSPI)obtained from Naver Finance from January 2000 to April 2020, the Oil Price datafor the entire Gyeongnam region among domestic oil price data obtained fromKorea Petroleum Corporation (Opinet) and USD Exchange data provided byNaver Financial from April 2004 to May 2020 were employed. The analyses weremade using mean absolute error (MAE), mean squared error (MSE) and rootmean squared error (RMSE) as the performance metric. Findings: Empiricalresults show that the proposed method outperforms all other models thatserve as benchmarked models, in all three kinds of data used in this research.This also shows that ensemble models have better performance than thesingle models as both AdaBoost-GRU and AdaBoost-LSTM outperform theirrespective single GRU and single LSTM. Novelty/Applications: This empiricalstudy suggests that the AdaBoost-GRU ensemble-learning model is a highlypromising approach for forecasting these kinds of data. However, anotherensemble model that can combine AdaBoost with other single models suchas ConvD1 can be developed and applied