Implementasi Model Fedformer Dalam Peramalan Harga Saham Harian Sektor Perbankan Di Indonesia: Studi Komparatif Dengan Lstm

Ocrisa, Joice (2026) Implementasi Model Fedformer Dalam Peramalan Harga Saham Harian Sektor Perbankan Di Indonesia: Studi Komparatif Dengan Lstm. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221066-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Sektor perbankan merupakan salah satu sektor strategis di pasar modal Indonesia yang memiliki peran penting dalam mencerminkan kondisi makroekonomi Indonesia. Oleh karena itu, peramalan harga saham sektor ini menjadi krusial bagi investor, regulator, dan pembuat kebijakan. Namun, kompleksitas data saham yang bersifat non-linier, dinamis, serta dipengaruhi oleh berbagai faktor eksternal membuat proses peramalan menjadi tantangan tersendiri. Penelitian ini bertujuan untuk mengimplementasikan dan mengevaluasi model Frequency Enhanced Decomposed Transformer (FEDformer) dalam melakukan peramalan harga saham harian sektor perbankan di Indonesia pada periode 30 Juni 2004 sampai 30 Juni 2025 serta membandingkannya dengan model Long Short-Term Memory (LSTM) sebagai model pembanding. Hasil penelitian menunjukkan bahwa model FEDformer menghasilkan kinerja prediksi yang lebih baik dibandingkan LSTM. Model FEDformer mampu mencapai RMSE sebesar 97,44 dan MAPE sebesar 1,59% untuk saham BBRI, sedangkan pada emiten yang sama model LSTM menghasilkan RMSE sebesar 170,01 dan MAPE sebesar 3,00%. Selain itu, untuk seluruh saham perbankan yang dianalisis, model FEDformer mampu menghasilkan nilai RMSE dan MAPE yang selalu lebih rendah LSTM. Berdasarkan hasil tersebut, dapat disimpulkan bahwa model FEDformer mampu menangkap pola musiman dan tren jangka panjang lebih efektif dibandingkan LSTM.
=====================================================================================================================================
The banking sector is one of the strategic sectors in the Indonesian capital market and plays an important role in reflecting Indonesia’s macroeconomic conditions. Therefore, forecasting stock prices in this sector is crucial for investors, regulators, and policymakers. However, the complexity of stock price data, which is nonlinear, dynamic, and influenced by various external factors, makes the forecasting process particularly challenging. This study aims to implement and evaluate the Frequency Enhanced Decomposed Transformer (FEDformer) model for forecasting daily stock prices in Indonesia’s banking sector over the period from June 30, 2004 to June 30, 2025, and to compare its performance with the Long Short-Term Memory (LSTM) model as a benchmark. The results show that the FEDformer model achieves superior predictive performance compared to LSTM. Specifically, FEDformer attains an RMSE of 97.44 and a MAPE of 1.59% for BBRI, whereas the LSTM model produces an RMSE of 170.01 and a MAPE of 3.00% for the same stock. Furthermore, across all banking stocks analyzed, FEDformer consistently yields lower RMSE and MAPE values than LSTM. Based on these findings, it can be concluded that the FEDformer model is more effective in capturing seasonal patterns and long-term trends compared to LSTM.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	FEDformer, LSTM, Peramalan Harga Saham, Sektor Perbankan, Time Series.
Subjects:	Q Science Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA404 Fourier series Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.F56 Data structures (Computer science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Joice Ocrisa
Date Deposited:	30 Jan 2026 10:19
Last Modified:	30 Jan 2026 10:19
URI:	http://repository.its.ac.id/id/eprint/131373

Actions (login required)

View Item