Analisis parameter long short-term memory dalam memprediksi ihsg (indeks harga saham gabungan) = Analysis of parameters of long short-term memory in predicting ihsg (composite stock price index)

Remetwa, Daud Padut Aritiran (2024) Analisis parameter long short-term memory dalam memprediksi ihsg (indeks harga saham gabungan) = Analysis of parameters of long short-term memory in predicting ihsg (composite stock price index). Bachelor thesis, Universitas Pelita Harapan.

	Text (Title) Title.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (51kB)
	Text (Abstract) Abstract.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (603kB)
	Text (ToC) ToC.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (686kB)
	Text (Chapter1) Chapter1.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (617kB)
	Text (Chapter2) Chapter2.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (779kB)
	Text (Chapter3) Chapter3.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (834kB)
	Text (Chapter4) Chapter4.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (936kB)
	Text (Chapter5) Chapter5.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (558kB)
	Text (Bibliography) Bibliography.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (581kB)
	Text (Appendices) Appendices.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (4MB)

Abstract

Investasi saham populer bagi investor yang ingin meningkatkan nilai asset finansial mereka, meskipun ada risiko signifikan. Untuk mengelola risiko dan memaksimalkan hasil, penelitian ini menggunakan model Long Short-Term Memory (LSTM) untuk memprediksi harga Indeks Harga Saham Gabungan (IHSG) di pasar modal Indonesia. Penelitian berfokus pada parameter kunci dalam model LSTM seperti jumlah sliding window, epoch, learning rate, dan jenis optimizer. Empat konfigurasi diuji: pertama, sliding window sebesar 5, epoch sebanyak 10, learning rate sebesar 0,001, dan optimizer Adam, (xv + 62 halaman; 14 gambar; 45 tabel; 4 lampiran) menghasilkan RMSE sebesar 7,2218. Setelah pengulangan 20 kali, rata-rata RMSE adalah 140,00 dan R-square sebesar 0,97, menunjukkan kombinasi ini kurang efektif. Kedua, sliding window sebesar 20, epoch sebanyak 40, learning rate sebesar 0,001, dan optimizer Adam menghasilkan RMSE sebesar 1,7272, dan setelah 20 kali pengulangan, rata-rata RMSE adalah 117,36 dan R-square sebesar 0,95, menunjukkan peningkatan akurasi tetapi masih belum optimal. Ketiga, sliding window sebesar 5, epoch sebanyak 50, learning rate sebesar 0,01, dan optimizer Adam menghasilkan rata-rata RMSE sebesar 100,92 dan R-square sebesar 0,96 setelah 20 kali pengulangan, menunjukkan peningkatan akurasi yang lebih baik. Keempat, sliding window sebesar 5, epoch sebanyak 50, learning rate sebesar 0,01, dan optimizer Adam menghasilkan performa terbaik dengan rata-rata RMSE sebesar 92,28 dan R-square sebesar 0,99 setelah 20 kali pengulangan, menunjukkan bahwa konfigurasi ini memiliki tingkat kesalahan terendah dan kemampuan prediksi yang paling akurat. Kesimpulannya, konfigurasi keempat dengan parameter sliding window sebesar 5, epoch sebanyak 50, learning rate sebesar 0,01, dan optimizer Adam memberikan performa terbaik dalam memprediksi harga IHSG, menunjukkan pentingnya pemilihan parameter yang tepat dalam meningkatkan akurasi prediksi harga saham menggunakan model LSTM./ Stock investment is popular among investors who want to increase the value of their financial assets, despite the significant risks involved. To manage risk and maximize returns, this study uses a Long Short-Term Memory (LSTM) model to predict the Indonesia Composite Index (IHSG) prices in the Indonesian stock market. The research focuses on key parameters in the LSTM model such as the number of sliding windows, epochs, learning rate, and type of optimizer. Four configurations were tested: first, a sliding window of 5, 10 epochs, a learning (xv + 62 pages, 14 figures, 45 tables; 4 appendices) rate of 0.001, and Adam optimizer, resulting in an RMSE of 7.2218. After 20 iterations, the average RMSE was 140.00 and the R-square was 0.97, indicating this combination was less effective. Second, a sliding window of 20, 40 epochs, a learning rate of 0.001, and Adam optimizer produced an RMSE of 1.7272, and after 20 iterations, the average RMSE was 117.36 and the R-square was 0.95, showing improved accuracy but still not optimal. Third, a sliding window of 5, 50 epochs, a learning rate of 0.01, and Adam optimizer resulted in an average RMSE of 100.92 and an R-square of 0.96 after 20 iterations, showing better accuracy. Fourth, a sliding window of 5, 50 epochs, a learning rate of 0.01, and Adam optimizer produced the best performance with an average RMSE of 92.28 and an R-square of 0.99 after 20 iterations, indicating that this configuration had the lowest error rate and the most accurate predictive capability. In conclusion, the fourth configuration with a sliding window of 5, 50 epochs, a learning rate of 0.01, and Adam optimizer provided the best performance in predicting IHSG prices, highlighting the importance of selecting the right parameters to improve the accuracy of stock price predictions using the LSTM model.

Actions (login required)

View Item