Pengembangan Model Prediksi Harga Saham Berbasis LSTM-GRU di Bursa Efek Indonesia

Hidayat, Bagus Febrian Dali (2025) Pengembangan Model Prediksi Harga Saham Berbasis LSTM-GRU di Bursa Efek Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025201208-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Pasar saham Indonesia, khususnya indeks LQ45, memiliki potensi tinggi namun juga volatilitas yang besar, sehingga diperlukan model prediksi harga yang akurat. Penelitian ini bertujuan mengembangkan model prediksi berbasis deep learning menggunakan pendekatan hybrid Long Short-Term Memory (LSTM) dan Gated Recurrent Unit (GRU). Data yang digunakan dalam penelitian ini berupa harga penutupan harian saham-saham LQ45 selama lima tahun terakhir. Data melalui tahap preprocessing yang mencakup penanganan missing values, normalisasi dengan MinMaxScaler, serta pembentukan dataset sekuensial dengan window size 30 hari. Data dibagi menjadi tiga subset: 70% untuk pelatihan, 15% untuk validasi, dan 15% untuk pengujian. Model dievaluasi menggunakan menggunakan metrik Mean Absolute Error (MAE) dan Root Mean Squared Error (RMSE). Hasil eksperimen menunjukkan bahwa model hybrid LSTM-GRU dengan konfigurasi 128–32 memberikan hasil prediksi terbaik dibandingkan model-model pembanding. Model ini mencatatkan MAE (%) sebesar 2.37 dan RMSE sebesar 103.08, mengungguli LSTM tunggal, GRU tunggal, SVM, dan Random Forest. Selain itu, performa model bervariasi antar sektor, dengan akurasi tertinggi pada sektor Konsumsi dan Farmasi & Kesehatan. Temuan ini memperkuat potensi penerapan model deep learning berbasis hybrid untuk prediksi harga saham di pasar modal Indonesia.
====================================================================================================================================
Indonesia's stock market, particularly the LQ45 index, offers high potential but also exhibits considerable volatility, requiring accurate stock price prediction models. This study aims to develop a deep learning-based prediction model using a hybrid approach combining Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The data used in this research consists of daily closing prices of LQ45 stocks over the past five years. The data underwent preprocessing, including handling of missing values, normalization using MinMaxScaler, and sequential dataset construction with a window size of 30 days. The dataset was split into three subsets: 70% for training, 15% for validation, and 15% for testing. Model performance was evaluated using the Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) metrics. Experimental results show that the hybrid LSTM-GRU model with a 128–32 configuration outperformed all other benchmark models. This model achieved an MAE (%) of 2.37 and an RMSE of 103.08, surpassing standalone LSTM, GRU, SVM, and Random Forest. Additionally, the model's performance varied across sectors, with the highest accuracy observed in the Consumer and Healthcare sectors. These findings reinforce the potential of hybrid deep learning models for stock price prediction in the Indonesian capital market.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	GRU, LSTM, Prediksi Harga Saham, Hybrid LSTM-GRU, Deep Learning, GRU, LSTM, Stock Price Prediction, Hybrid LSTM–GRU, Deep Learning
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Bagus Febrian Dali Hidayat
Date Deposited:	23 Jul 2025 07:07
Last Modified:	23 Jul 2025 07:07
URI:	http://repository.its.ac.id/id/eprint/120890

Actions (login required)

View Item