Peramalan data time series menggunakan model hybrid ARIMA-XGBoost

Anggandani, Nizar Zaki (2026) Peramalan data time series menggunakan model hybrid ARIMA-XGBoost. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

Preview	Text 1_cover.pdf Download (419kB) | Preview
Preview	Text 2_abstrak.pdf Download (175kB) | Preview
Preview	Text 3_skbebasplagiarism.pdf Download (394kB) | Preview
Preview	Text 4_daftarisi.pdf Download (175kB) | Preview
Preview	Text 5_bab1.pdf Download (163kB) | Preview
	Text 6_bab2.pdf Restricted to Registered users only Download (3MB) | Request a copy
	Text 7_bab3.pdf Restricted to Registered users only Download (750kB) | Request a copy
	Text 8_bab4.pdf Restricted to Registered users only Download (836kB) | Request a copy
	Text 9_bab5.pdf Restricted to Registered users only Download (162kB) | Request a copy
	Text 10_daftarpustaka.pdf Restricted to Registered users only Download (153kB) | Request a copy
	Text 11_lampiran.pdf Restricted to Repository staff only Download (362kB) | Request a copy

Abstract

Peramalan data time series berperan penting dalam mendukung pengambilan keputusan, termasuk pada sektor transportasi udara. Namun, data time series umumnya mengandung pola linier dan nonlinier sehingga penggunaan model tunggal sering kali belum mampu merepresentasikan seluruh karakteristik data secara optimal. Oleh karena itu, penelitian ini bertujuan menerapkan model hybrid ARIMA-XGBoost pada data jumlah kedatangan penumpang pesawat internasional di Bandara Soekarno-Hatta. Model hybrid dibangun dengan menggunakan ARIMA untuk memodelkan komponen linier, sedangkan residual ARIMA digunakan sebagai masukan XGBoost untuk memodelkan komponen nonlinier yang masih tersisa. Kinerja model ARIMA, XGBoost, dan hybrid ARIMA-XGBoost dibandingkan pada horizon peramalan 1-step ahead, 3-step ahead, dan 5-step ahead. Optimasi hyperparameter XGBoost dilakukan menggunakan Grid Search 5-Fold Time Series Cross Validation, sedangkan evaluasi model menggunakan RMSE, MAE, dan MAPE. Hasil penelitian menunjukkan bahwa horizon terbaik berbeda pada setiap model, yaitu ARIMA pada horizon 5-step ahead, sedangkan XGBoost dan hybrid ARIMA-XGBoost pada horizon 1-step ahead. Pada horizon terbaik model hybrid, diperoleh RMSE sebesar 69423.12, MAE sebesar 52075.29, dan MAPE sebesar 7.31%, yang lebih rendah dibandingkan model ARIMA dan XGBoost. Hasil tersebut menunjukkan bahwa pendekatan hybrid ARIMA-XGBoost memberikan kinerja peramalan terbaik pada data yang digunakan dalam penelitian ini. Time series forecasting is an essential tool for supporting decision-making across various sectors, including air transportation. However, time series data generally exhibit both linear and nonlinear patterns, making a single forecasting model insufficient to fully capture their underlying characteristics. Therefore, this study aims to apply the hybrid ARIMA–XGBoost model to forecast the number of international passenger arrivals at Soekarno-Hatta International Airport. The hybrid model employs ARIMA to model the linear component, while the residuals generated by ARIMA are used as input for XGBoost to capture the remaining nonlinear patterns. The forecasting performance of ARIMA, XGBoost, and the hybrid ARIMA-XGBoost model is compared using 1-step ahead, 3-step ahead, and 5-step ahead forecasting horizons. XGBoost hyperparameters are optimized using Grid Search with 5-Fold Time Series Cross Validation, and model performance is evaluated using RMSE, MAE, and MAPE. The results indicate that the optimal forecasting horizon differs across models: ARIMA performs best at the 5-step ahead horizon, whereas XGBoost and the hybrid ARIMA-XGBoost model achieve their best performance at the 1-step ahead horizon. At this horizon, the hybrid model achieves an RMSE of 69423.12, an MAE of 52075.29, and a MAPE of 7.31%, outperforming both ARIMA and XGBoost. These findings demonstrate that the hybrid ARIMA-XGBoost model provides superior forecasting performance for the data used in this study.

Actions (login required)

View Item