Peramalan Harga Saham Perbankan Menggunakan Metode Hybrid ARIMA-LSTM Dengan Pendekatan Discrete Wavelet Transformation

Hidayah, Dian Nauroh (2025) Peramalan Harga Saham Perbankan Menggunakan Metode Hybrid ARIMA-LSTM Dengan Pendekatan Discrete Wavelet Transformation. Other thesis, INSTITUT TEKNOLOGI SEPULUH NOPEMBER.

Text 5006211008-Undergraduate Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Sektor perbankan memegang peranan krusial dalam perekonomian Indonesia, dengan saham-saham seperti BBCA, BBRI, dan BMRI menjadi kontributor utama kapitalisasi pasar. Tingginya volatilitas harga saham yang dipengaruhi berbagai faktor eksternal dan internal menuntut adanya metode prediksi yang akurat untuk meminimalkan risiko investasi. Penelitian ini mengembangkan model hybrid ARIMA-LSTM yang dikombinasikan dengan Discrete Wavelet Transformation (DWT) untuk meningkatkan akurasi prediksi harga saham. DWT digunakan untuk mendekomposisi data harga saham harian periode 2022–2024 menjadi komponen linear (approximation) yang dianalisis menggunakan ARIMA dan komponen non-linear (detail) yang diproses dengan LSTM. Hasil penelitian menunjukkan bahwa model ARIMA terbaik untuk masing-masing saham adalah BBCA dengan ARIMA([43],1,[29]), BBRI dengan ARIMA([24,42],1,0), dan BMRI dengan ARIMA([6],1,[20,56]). Sementara itu, LSTM dioptimalkan menggunakan kombinasi hyperparameter spesifik: BBCA (neurons 30, dropout 0,1, epochs 200), BBRI (neurons 50, dropout 0,1, epochs 200), dan BMRI (neurons 30, dropout 0,1, epochs 100). Setelah dilakukan proses rekonstruksi sinyal, model hybrid ini menghasilkan tingkat akurasi tertinggi pada saham BMRI dengan MAPE sebesar 6,8%, diikuti oleh BBCA sebesar 8,5%, dan BBRI sebesar 17%. Hasil ini mengindikasikan bahwa pendekatan hybrid ARIMA-LSTM berbasis DWT efektif untuk memprediksi harga saham.
============================================================================================================================================
The banking sector plays a crucial role in Indonesia’s economy, with stocks such as BBCA, BBRI, and BMRI being major contributors to market capitalization. The high volatility of stock prices, influenced by various external and internal factors, demands accurate forecasting methods to minimize investment risks. This study develops a hybrid ARIMA-LSTM model integrated with Discrete Wavelet Transformation (DWT) to improve the accuracy of stock price prediction. DWT is employed to decompose daily stock price data from 2022 to 2024 into linear components (approximation) analyzed using ARIMA, and non-linear components (detail) processed using LSTM. The results indicate that the best ARIMA models for each stock are BBCA with ARIMA([43],1,[29]), BBRI with ARIMA([24,42],1,0), and BMRI with ARIMA([6],1,[20,56]). Meanwhile, the LSTM models were optimized with specific hyperparameter combinations: BBCA (neurons 30, dropout 0,1, epochs 200), BBRI (neurons 50, dropout 0,1, epochs 200), and BMRI (neurons 30, dropout 0,1, epochs 100). After signal reconstruction, the hybrid model achieved the highest accuracy for BMRI with a MAPE of 6,8%, followed by BBCA at 8,5%, and BBRI at 17%. These results indicate that the DWT-based hybrid ARIMA-LSTM approach is effective in forecasting stock prices

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Autoregressive Integrated Moving Average, Discrete Wevelet Transformation, Long Short Term Mermories, Saham LQ45. Autoregressive Integrated Moving Average, Discrete Wevelet Transformation, Long Short Term Mermories, LQ45 Index
Subjects:	H Social Sciences > HG Finance > HG4529 Investment analysis H Social Sciences > HG Finance > HG4915 Stocks--Prices
Divisions:	Faculty of Mathematics, Computation, and Data Science > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Dian Nauroh Hidayah
Date Deposited:	01 Aug 2025 01:07
Last Modified:	01 Aug 2025 01:10
URI:	http://repository.its.ac.id/id/eprint/125227

Actions (login required)

View Item