Peramalan Jumlah Penumpang Kereta Api Nasional dengan Model Hybrid ARIMAX-XGBoost Menggunakan Dummy Intervensi

Authors

  • Rafi Kamindra Universitas Pendidikan Indonesia
  • Dadan Dasari Universitas Pendidikan Indonesia
  • Fitriani Agustina Universitas Pendidikan Indonesia

DOI:

https://doi.org/10.14421/fourier.2026.151.45-53

Keywords:

ARIMAX, Jumlah Penumpang Kereta Api, Model Hybrid, Peramalan, Variabel Dummy Intervensi, XGBoost

Abstract

Abstrak

Keberadaan guncangan struktural (structural break) pada data deret waktu menyebabkan metode peramalan linear konvensional memiliki keterbatasan dalam menghasilkan akurasi yang tinggi. Keterbatasan ini sangat terlihat pada data mobilitas penumpang kereta api nasional yang merekam fluktuasi ekstrem dan perubahan level secara permanen akibat krisis makro di masa lalu, seperti pandemi COVID-19. Penelitian ini bertujuan membangun model peramalan yang adaptif menggunakan pendekatan Hybrid yang menggabungkan model statistik linear Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) dan algoritma Machine Learning non-linear eXtreme Gradient Boosting (XGBoost). Metodologi dilakukan dalam dua tahap, yaitu estimasi komponen linear dan dampak kebijakan menggunakan ARIMAX dengan empat variabel dummy intervensi, dilanjutkan dengan pemodelan sisaan (residual) yang mengandung pola non-linear menggunakan XGBoost. Hasil analisis menunjukkan model linear terbaik adalah ARIMAX(0,1,3) dengan residu yang terbukti tidak berdistribusi normal, menjustifikasi penggunaan XGBoost pada tahap kedua. Evaluasi kinerja pada data uji (out-of-sample) membuktikan bahwa model Hybrid ARIMAX-XGBoost lebih unggul dibandingkan model ARIMAX tunggal, dengan penurunan nilai RMSE sebesar 2,76% dan perbaikan akurasi MAPE sebesar 2,81%, sehingga pendekatan ini lebih efektif digunakan untuk perencanaan operasional transportasi di era pemulihan pasca-pandemi.

Abstract 

The existence of structural breaks in time series data causes conventional linear forecasting methods to have limitations in producing high accuracy. This limitation is highly evident in the national train passenger mobility data, which records extreme fluctuations and permanent level shifts due to past macro crises, such as the COVID-19 pandemic. This study aims to build an adaptive forecasting model using a Hybrid approach combining the linear statistical model Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) and the non-linear Machine Learning algorithm eXtreme Gradient Boosting (XGBoost). The methodology was conducted in two stages, estimating linear components and policy impacts using ARIMAX with four intervention dummy variables, followed by modeling the residuals containing non-linear patterns using XGBoost. The analysis results indicated that the best linear model was ARIMAX(0,1,3) with residuals proven to be non-normally distributed, justifying the use of XGBoost in the second stage. Performance evaluation on testing data (out-of-sample) demonstrated that the Hybrid ARIMAX-XGBoost model outperformed the single ARIMAX model, with a 2.76% decrease in RMSE and a 2.81% improvement in MAPE accuracy, proving that this approach is more effective for transportation operational planning in the post-pandemic recovery era.

Downloads

Download data is not yet available.

References

A. Leliana, S. Ferlianne, and M. Muhardjito, "Optimalisasi jumlah perjalanan KRL lintas Jakarta Kota-Bogor terhadap demand penumpang menggunakan ARIMA Box-Jenkins," Journal of Sustainable Civil Engineering (JOSCE), vol. 5, no. 1, 2023.

K. Sugiono and M. Herlina, "Analisis fuzzy time series pada jumlah penumpang kereta api di wilayah Sumatera menggunakan metode Markov chain dan Lee," Bandung Conference Series: Statistics, vol. 4, no. 1, pp. 291-301, 2024.

Y. Angraini, A. Pangestika, and I. Sumertajaya, "Comparison of the symmetric and asymmetric generalized autoregressive conditional heteroscedasticity (GARCH) models in forecasting the 2018-2023 Jakarta Composite Index," ComTech, vol. 15, no. 1, pp. 1-15, 2024.

M. Arumsari and A. Dani, "Peramalan data runtun waktu menggunakan model hybrid time series regression autoregressive integrated moving average," Jurnal Siger Matematika, vol. 2, no. 1, 2021.

W. W. S. Wei, Time series analysis: Univariate and multivariate methods, 2nd ed. Pearson Education, 2006.

A. Hadianfar, R. Yousefi, M. Delavary, V. Fakoor, M. Shakeri, and M. Lavallière, "Effects of government policies and the Nowruz holidays on confirmed COVID-19 cases in Iran: An intervention time series analysis," PLOS One, vol. 16, no. 8, p. e0256516, 2021.

L. Huang, X. Zhao, Y. Liu, and P. Yang, "Analysis of the atmospheric duct existence factors in tropical cyclones based on the SHAP interpretation of extreme gradient boosting predictions," Remote Sensing, vol. 14, no. 16, p. 3952, 2022.

S. Odhiambo, N. Cornelious, and H. Waititu, "Developing a hybrid ARIMA-XGBoost model for analysing mobile money transaction data in Kenya," Asian Journal of Probability and Statistics, vol. 26, no. 10, pp. 108-126, 2024.

Badan Pusat Statistik, "Penumpang Angkutan Kereta Api Nasional Bulanan," Badan Pusat Statistik. [Daring]. Tersedia: https://www.bps.go.id/id/statistics-table/2/MjM1MyMy/penumpang-angkutan-kereta-api-nasional-bulanan-.html (Diakses: 24 September 2025).

V. R. Joseph, "Optimal ratio for data splitting," Statistical Analysis and Data Mining: The ASA Data Science Journal, vol. 15, no. 4, pp. 531-538, 2022.

G. E. P. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time series analysis: Forecasting and control, 5th ed. John Wiley & Sons, 2015.

I. Benitez, J. Ibañez, C. Lumabad, J. Cañete, and J. Principe, "Day-ahead hourly solar photovoltaic output forecasting using SARIMAX, long short-term memory, and extreme gradient boosting: Case of the Philippines," Energies, vol. 16, no. 23, p. 7823, 2023.

K. Sivhugwana and E. Ranganai, "An ensemble approach to short-term wind speed predictions using stochastic methods, wavelets and gradient boosting decision trees," Wind, vol. 4, no. 1, pp. 44-67, 2024.

H. Hassani, M. Royer-Carenzi, L. Mashhad, M. Yarmohammadi, and M. Yeganegi, "Exploring the depths of the autocorrelation function: Its departure from normality," Information, vol. 15, no. 8, p. 449, 2024.

N. Al-Saati, I. Omran, A. Al-Taai, Z. Al-Saati, and K. Hashim, "Statistical modeling of monthly streamflow using time series and artificial neural network models: Hindiya barrage as a case study," Water Practice & Technology, vol. 16, no. 2, pp. 681-691, 2021.

Downloads

Published

2026-04-30

How to Cite

Kamindra, R. ., Dasari, D., & Agustina, F. (2026). Peramalan Jumlah Penumpang Kereta Api Nasional dengan Model Hybrid ARIMAX-XGBoost Menggunakan Dummy Intervensi. Jurnal Fourier, 15(1), 45–53. https://doi.org/10.14421/fourier.2026.151.45-53

Issue

Section

Articles