Journal Article

Extreme gradient boosting (XGBoost) method in making forecasting application and analysis of USD exchange rates against rupiah

Sholahuddin, A, Islam, S F N, Abdullah, A S

2021

Overview

Economic conditions in Indonesia are still unstable, causing the US dollar exchange rate to increase. This is because most international transactions in Indonesia use US dollars. Prediction or forecasting is chosen as one of the important things in choosing a market to invest in buying and selling. This research will focus on making forecasting applications and analyzing the exchange rate of USD against rupiah based on time series data or temporal datasets from the Investing.com site using machine learning methods, namely Extreme Gradient Boosting (XGBoost). Applications created using the python programming language and streamlit framework. Modeling is carried out using the Knowledge Discovery in Database (KDD) methodology with the stages of dividing the dataset with a 50:50 percentage share into test and train data. The modeling uses hyperparameter tuning values, namely n_estimators = 1000, max_depth = 1, x_colsample_bytree = 0.9894, x_gamma = 0.9989, x_min_child = 1.0, x_reg_lamda = 0.2381, and x_subsample = 0.7063 with best loss or RMSE 451.4151. The values of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) when making the model were 6.61374% and 3.95485%. Meanwhile, when testing the model, the RMSE is 0.23577% and MAPE is 0.11643%.

• Publisher: IOP Publishing
• Subject: Datasets / Economic analysis / Economic conditions / Economic forecasting / exchange rate / Extreme Gradient Boosting (XGBoost) / forecasting / Foreign exchange rates / Knowledge Discovery in Database (KDD) / Machine learning / Mathematical models / Mean Absolute Percentage Error (MAPE) / Model testing / Modelling / Physics / Programming languages / Root Mean Square Error (RMSE) / Root-mean-square errors / streamlit framework