PREDIKSI HARGA SAHAM MENGGUNAKAN METODE ETSFORMER STUDI KASUS: PTBA

Muhammad Azka Atqiya, - and Lala Septem Riza, - and Ani Anisyah, - (2025) PREDIKSI HARGA SAHAM MENGGUNAKAN METODE ETSFORMER STUDI KASUS: PTBA. S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Prediksi harga saham pada sektor komoditas volatil seperti PTBA merupakan tantangan signifikan karena karakteristik data yang non-linear dan non-stasioner. Arsitektur deep learning hibrida seperti ETSFormer telah muncul untuk mengatasi keterbatasan ini dengan mengintegrasikan dekomposisi deret waktu, namun efektivitasnya pada data saham komoditas di pasar negara berkembang belum banyak teruji sehingga penelitian ini bertujuan untuk mengaplikasikan dan mengevaluasi kinerja model ETSFormer dalam memprediksi harga saham PTBA sebagai studi kasus yang relevan. Hasil penelitian ini menjelaskan bahwa model ETSFormer terbaik, yang dioptimalkan melalui Grid Search dengan menguji kombinasi dimensi model, learning rate, dan batch size, ditentukan setelah evaluasi akhir pada tiga kandidat teratas. Hasil pengujian menunjukkan bahwa model final dengan konfigurasi hyperparameter dimensi model = 16, batch size = 16, dan learning rate = 0,1 berhasil mencapai kinerja prediksi yang sangat baik dengan nilai MAPE 3,23%, MAE 76,26, dan RMSE 106,14. Analisis lebih lanjut juga mengonfirmasi bahwa penyertaan fitur harga batubara harian secara signifikan meningkatkan akurasi, dan model menunjukkan keandalan tertinggi pada skenario prediksi jangka pendek. Predicting stock prices in volatile commodity sectors such as PTBA presents a significant challenge due to the non-linear and non-stationary characteristics of the data. Hybrid deep learning architectures such as ETSFormer have emerged to address these limitations by integrating time series decomposition; however, their effectiveness on commodity stock data in emerging markets has not been extensively tested. Therefore, this study aims to apply and evaluate the performance of the ETSFormer model in predicting PTBA stock prices as a relevant case study. The findings of this research indicate that the best ETSFormer model, optimized through Grid Search by testing combinations of model dimensions, learning rates, and batch sizes, was determined after final evaluation of the top three candidates. The test results show that the final model, with a hyperparameter configuration of model dimension = 16, batch size = 16, and learning rate = 0.1, achieved excellent predictive performance with MAPE of 3.23%, MAE of 76,26, and RMSE of 106.14. Further analysis also confirmed that incorporating daily coal price features significantly improved accuracy, and the model demonstrated the highest reliability in short-term prediction scenarios.

	Text S_KOM_2100812_Title.pdf Download (677kB)
	Text S_KOM_2100812_Chapter1.pdf Download (79kB)
	Text S_KOM_2100812_Chapter2.pdf Restricted to Staf Perpustakaan Download (767kB)
	Text S_KOM_2100812_Chapter3.pdf Download (106kB)
	Text S_KOM_2100812_Chapter4.pdf Restricted to Staf Perpustakaan Download (1MB)
	Text S_KOM_2100812_Chapter5.pdf Download (57kB)
	Text S_KOM_2100812_Appendix.pdf Restricted to Staf Perpustakaan Download (126kB)

Item Type:	Thesis (S1)
Additional Information:	https://scholar.google.com/citations?hl=en&user=BSCar9oAAAAJ&view_op=list_works&sortby=title ID SINTA Dosen Pembimbing: Lala Septem Riza: 5975668 Ani Anisyah: 6786982
Uncontrolled Keywords:	ETSFormer, Grid Search, hyperparameter, saham, prediksi. ETSFormer, Grid Search, hyperparameter, prediction, stock.
Subjects:	L Education > L Education (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Fakultas Pendidikan Matematika dan Ilmu Pengetahuan Alam > Program Studi Ilmu Komputer
Depositing User:	Muhammad Azka Atqiya
Date Deposited:	09 Sep 2025 07:21
Last Modified:	09 Sep 2025 07:21
URI:	http://repository.upi.edu/id/eprint/138088

Actions (login required)

View Item