Salma, Talitha (2026) Analisis Perbandingan Prediksi IHSG Menggunakan Gated Recurrent Unit Berbasis Analisis Sentimen Berdasarkan Pendekatan Valence Aware Dictionary and sEntiment Reasoner. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5006221013-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5006221013-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (11MB) | Request a copy |
Abstract
Indeks Harga Saham Gabungan (IHSG) merupakan indikator utama untuk menilai kinerja pasar saham secara keseluruhan sehingga memiliki peran penting dalam perekonomian dan menjadi acuan bagi investor dalam pengambilan keputusan. Pergerakan IHSG yang fluktuatif dipengaruhi berbagai faktor, termasuk sentimen. Penelitian ini bertujuan untuk memperoleh hasil analisis sentimen IHSG menggunakan metode Valence Aware Dictionary and sEntiment Reasoner (VADER), mengetahui karakteristik data sentimen dan harga penutupan IHSG, membandingkan kinerja model Gated Recurrent Unit (GRU) tanpa dan dengan variabel sentimen, serta menghasilkan peramalan harga penutupan IHSG untuk lima periode selanjutnya. Data yang digunakan berupa data harian mencakup data teknikal, harga penutupan IHSG, dan data sentimen yang diperoleh melalui crawling dari platform X pada periode 1 Januari 2022 hingga 31 Agustus 2025. Analisis sentimen dilakukan menggunakan VADER, metode berbasis leksikon yang mengukur polaritas positif, negatif, dan netral. Prediksi dilakukan menggunakan GRU yang efektif dalam memproses data sekuensial dengan arsitektur lebih sederhana dibandingkan LSTM. Hasil analisis menunjukkan bahwa tweet terkait IHSG didominasi sentimen positif sebesar 47,875%, diikuti netral 31,057% dan negatif 21,068%. Hal ini menunjukkan opini publik yang cenderung optimistis, meskipun sebagian bersifat informatif atau mengungkapkan kekhawatiran. Pemodelan GRU memberikan nilai MAPE <10% pada kedua model, menandakan kemampuan prediksi sangat baik. Model dengan sentimen menghasilkan performa terbaik dengan MAPE 0,623%, sedikit lebih unggul dibandingkan model tanpa sentimen sebesar 0,634%. Oleh karena itu, model dengan sentimen dipilih sebagai model optimal dan digunakan untuk proses peramalan. Model terbaik didapat dengan konfigurasi hyperparameter optimizer Adam, timestep 10, neuron 50, dropout 0,02, batch size 32, learning rate 0,01, epoch 16 melalui early stopping, dan replikasi ke-3. Hasil peramalan lima hari ke depan menunjukkan pola historis koreksi jangka pendek. Hasil penelitian ini diharapkan dapat memberikan wawasan mengenai pengaruh sentimen pasar dan mendukung proses pengambilan keputusan investasi.
====================================================================================================================================
The Indonesia Stock Exchange Composite Index (IHSG) is a key indicator for assessing overall stock market performance, thus playing a crucial role in the economy and serving as a reference for investors in decision-making. The fluctuating IHSG movement is influenced by various factors, including sentiment. This study aims to obtain IHSG sentiment analysis results using the Valence Aware Dictionary and Sentiment Reasoner (VADER) method, determine the characteristics of sentiment data and IHSG closing prices, compare the performance of the Gated Recurrent Unit (GRU) model with and without sentiment variables, and generate IHSG closing price projections for the next five periods. The data used are daily data including technical data, IHSG closing prices, and sentiment data obtained through crawling from platform X for the period January 1, 2022, to August 31, 2025. Sentiment analysis was performed using VADER, a lexicon-based method that measures positive, negative, and neutral polarities. Predictions are made using GRU, which is effective in processing sequential data with a simpler architecture than LSTM. The analysis showed that tweets related to the IHSG were dominated by positive sentiment at 47,875%, followed by neutral sentiment at 31,057% and negative sentiment at 21,068%. This indicates a tendency for public opinion to be optimistic, although some were informative or expressed concerns. GRU modeling provided MAPE values <10% for both models, indicating excellent predictive ability. The model with sentiment performed best with a MAPE of 0,623%, slightly superior to the model without sentiment at 0,634%. Therefore, the model with sentiment was selected as the optimal model and used for the forecasting process. The best model was configured with the Adam optimizer, timestep 10, neuron 50, dropout 0,02, batch size 32, learning rate 0,01, 16 epochs with early stopping, and 3 replications. The five-day forecast results showed a historical pattern of short-term corrections. The results of this study are expected to provide insight into the influence of market sentiment and support the investment decision-making process.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | GRU, Indeks Harga Saham Gabungan (IHSG), Prediksi, Sentimen, VADER. GRU, Indonesia Composite Stock Price Index (IHSG), Prediction, Sentiment, VADER. |
| Subjects: | Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) Q Science > QA Mathematics > QA76.9.I58 Recommender systems (Information filtering) |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis |
| Depositing User: | Talitha Salma Salma |
| Date Deposited: | 12 Jan 2026 07:03 |
| Last Modified: | 12 Jan 2026 07:03 |
| URI: | http://repository.its.ac.id/id/eprint/129495 |
Actions (login required)
 |
View Item |