Peramalan Jumlah Penumpang Kereta Api di Pulau Jawa Menggunakan Model BATS dan Model TBATS
Abstract
Abstract. Time series forecasting aims to find patterns in historical data series and extrapolate these patterns into the future. The success of a forecast depends on the use of forecasting methods, which must be adapted to the available data. Time series data that contains several seasonalities or has a dual-calendar effect is called data with complex seasonality, but not all forecasting methods can handle seasonal complexity. The methods that are able to handle seasonal complexity are the BATS (Box-Cox transform, ARMA errors, Trend, and Seasonal components) and TBATS (Trigonometric, Box-Cox transform, ARMA errors, Trend, and Seasonal components) models. Based on the results of the analysis carried out, it was found that the BATS model was better at predicting the number of train passengers on the island of Java compared to the TBATS model, as evidenced by the AIC value of 4318,364 and the RMSE of 1875,887 which were smaller than the AIC and RMSE values of the TBATS model. The results of forecasting using the BATS model showed that the highest number of train passengers was in December, namely 31,107,740 people and the lowest was in February 27,959,460 people.
Abstrak. Peramalan deret waktu bertujuan untuk menemukan pola dalam deret data historis dan mengekstrapolasikan pola tersebut ke masa depan. Keberhasilan suatu peramalan bergantung pada penggunaan metode peramalan yang harus disesuaikan dengan data yang tersedia. Data deret waktu yang mengandung beberapa musiman atau terdapat efek dual-calendar disebut data dengan musiman yang kompleks, namun tidak semua metode peramalan dapat menangani kompleksitas musiman. Adapun metode yang mampu menangani kompleksitas musiman adalah model BATS (Box-Cox transform, ARMA errors, Trend, and Seasonal components) dan TBATS (Trigonometric, Box-Cox transform, ARMA errors, Trend, and Seasonal components). Berdasarkan hasil analisis yang dilakukan didapati hasil bahwa model BATS lebih baik dalam meramalkan jumlah penumpang kereta api di pulau jawa dibandingkan dengan model TBATS dibuktikan dengan nilai AIC sebesar 4318.364 dan RMSE sebesar 1875.887 yang lebih kecil dibandingkan dengan nilai AIC dan RMSE dari model TBATS. Adapun hasil peramalan menggunakan model BATS didapat bahwa jumlah penumpang kereta api tertinggi adalah pada bulan Desember yaitu sebanyak 31.107.740 orang dan terendah pada bulan Februari 27.959.460 orang.
References
Hyndman, R. J., & Athanasopoulus. (2018). Forecasting : Principles and Practice. Australia: Monash University.
Baek, M. (2010). Forecasting Hourly Electricity Loads of South Korea: Innovations State Space Modeling Approach. The Korean Journal of Economics, 17.
De Livera, A., & Hyndman, R. (2009). Forecasting Time Series With Complex Seasonal Patterns Using Exponential Smoothing.
De Livera, A. M., Hyndman, R. J., & Snyder, R. D. (2010). Forecasting Time Series With Complex Seasonal Patterns using Exponential Smoothing.
Fhilip Trisa Parera, & Suwanda. (2023). Model Random Coefficient Autoregressive Orde Pertama dan Penerapannya. Jurnal Riset Statistika, 155–162. https://doi.org/10.29313/jrs.v3i2.3048
Jasmine Wildani Arisa, & Nusar Hajarisman. (2023). Determinasi Derajat Kelangsungan Hidup Anak Menggunakan Multigroup Structural Equation Modeling. Jurnal Riset Statistika, 147–154. https://doi.org/10.29313/jrs.v3i2.3047
Sahwa Chanigia Viqri, Z., & Kurniati, E. (2023). Perbandingan Penerapan Metode Fuzzy Time Series Model Chen-Hsu dan Model Lee dalam Memprediksi Kurs Rupiah terhadap Dolar Amerika. 1(1), 19–26. https://doi.org/10.29313/datamath.v1i1.12