Time Series Forecasting Using Ensemble Model

Abstract

Forecasting financial and economic time series has never been easy due to their sensitivity to political, economic, and social factors. Because of this, individuals who speculate throughout financial markets typically seek out robust models. Several intelligent algorithms have been proposed to forecast the financial market. The aim of this paper is to suggest a machine learning technique that combines linear regression as a base estimator with adaptive boosting regression. This combination is used to forecast future closing prices of gold commodity, NASDAQ, euro against the United States dollar, and British pound sterling against the United States dollar. In our approach, seven technical indicators are used for the data during the training of the adaptive boosting regression to improve the accuracy. Which will attempt to predict future closing prices. The four metrics root mean squared error, mean absolute percentage error, mean squared error and coefficient of determination measures for comparing several machine-learning models. To evaluate the effectiveness of different prediction models, all four metrics were computed. Experiment analysis proves that our approach provides better accuracy as compared to linear regression, adaboost regression and xgboost regression.