Prediksi Laju Pertumbuhan Produk Domestik Bruto (PDB) Menggunakan Hibrida ARIMAX-XGBoost Dengan Optimasi Optuna

Gustrifa, Wanda (2026) Prediksi Laju Pertumbuhan Produk Domestik Bruto (PDB) Menggunakan Hibrida ARIMAX-XGBoost Dengan Optimasi Optuna. Undergraduate thesis, UPN Veteran Jawa Timur.

[img] Text (Cover)
SkripsiCover.pdf

Download (705kB)
[img] Text (Bab 1)
Skripsi_Bab 1_organized.pdf

Download (94kB)
[img] Text (Bab 2)
Skripsi_Bab 2.pdf
Restricted to Repository staff only until 13 November 2028.

Download (637kB)
[img] Text (Bab 3)
Skripsi_Bab 3_organized.pdf
Restricted to Repository staff only until 13 November 2028.

Download (437kB)
[img] Text (Bab 4)
Skripsi_Bab4_organized.pdf
Restricted to Repository staff only until 13 November 2028.

Download (3MB)
[img] Text (Bab 5)
Skripsi_bab5.pdf

Download (76kB)
[img] Text (Daftar pustaka)
Skripsi_daftar pustaka.pdf

Download (101kB)

Abstract

Gross Domestic Product (GDP) growth is an important indicator of economic development. This study aims to compare the performance of ARIMA, ARIMAX, Hybrid ARIMAX–XGBoost, and Hybrid ARIMAX–XGBoost with Optuna optimization in forecasting Indonesia’s real GDP growth over one- to four-quarter horizons. The data consist of quarterly real GDP and export and import data for the 2000–2025 period, obtained from Statistics Indonesia (BPS). The data were transformed into year-on-year (YoY) growth rates, while the lags of the exogenous variables were determined using the Cross-Correlation Function (CCF). Historical shock periods were retained in the dataset and represented using a Dummy_Shock variable. The best ARIMAX specification obtained was ARIMAX(1,0,0). ARIMAX residuals were subsequently generated in an out-of-fold manner using an expanding-window approach and modeled with XGBoost under a direct multi-horizon forecasting strategy. Hyperparameter optimization was performed using Optuna with 200 trials for each forecasting horizon. Out-of-sample evaluation for the 2019–2025 period showed that the Hybrid ARIMAX–XGBoost+Optuna model achieved the lowest RMSE values of 1.9352, 2.7212, 3.1221, and 3.3096 for h=1 through h=4, respectively, and also produced the lowest Theil’s U across all horizons. Compared with ARIMA, RMSE decreased by 15.05%, 9.81%, 13.86%, and 20.60%, respectively. However, the largest performance improvement was obtained after incorporating exogenous information through ARIMAX, while XGBoost and Optuna provided additional gains. Based on MAE, ARIMAX performed best at h=1, whereas Hybrid ARIMAX–XGBoost achieved the best performance at h=2 through h=4. Therefore, Hybrid ARIMAX–XGBoost+Optuna demonstrated the most consistent performance in terms of RMSE and Theil’s U, although its superiority depended on the evaluation metric and forecasting horizon.

Item Type: Thesis (Undergraduate)
Contributors:
ContributionContributorsNIDN/NIDKEmail
Thesis advisorNugroho, BudiNIDN0707098003budinugroho.if@upnjatim.ac.id
Thesis advisorJunaidi, AchmadNIDN0710117803achmadjunaidi.if@upnjatim.ac.id
Subjects: T Technology > T Technology (General) > T58.6-58.62 Management Information Systems
Divisions: Faculty of Computer Science > Departemen of Informatics
Depositing User: Unnamed user with email 21081010083@student.upnjatim.ac.id
Date Deposited: 13 Nov 2026 07:28
Last Modified: 13 Nov 2026 07:28
URI: https://repository.upnjatim.ac.id/id/eprint/61342

Actions (login required)

View Item View Item