Anggraini, Novita (2025) IMPLEMENTASI SPATIAL TEMPORAL ATTENTION-BASED CONVOLUTIONAL NETWORK UNTUK PREDIKSI INDEKS HARGA SAHAM MENGGUNAKAN DATA TEKS DAN NUMERIK. Undergraduate thesis, Universitas Pembangunan Nasional Veteran Jawa Timur.
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Cover)
Cover.pdf Download (2MB) |
|
Text (Bab I)
21083010104_Bab I.pdf Download (235kB) |
|
Text (Bab II)
21083010104_Bab II.pdf Restricted to Repository staff only until 17 March 2028. Download (773kB) | Request a copy |
|
Text (Bab III)
21083010104_Bab III.pdf Restricted to Repository staff only until 17 March 2028. Download (492kB) | Request a copy |
|
Text (Bab IV)
21083010104_Bab IV.pdf Restricted to Repository staff only until 17 March 2028. Download (3MB) | Request a copy |
|
Text (Bab V)
21083010104_Bab V.pdf Restricted to Repository staff only Download (194kB) | Request a copy |
|
Text (Daftar Pustaka)
21083010104_Daftar Pustaka.pdf Download (244kB) |
|
Text (Lampiran)
21083010104_Lampiran.pdf Restricted to Repository staff only until 17 March 2028. Download (659kB) | Request a copy |
Abstract
Stock price fluctuations are influenced by various factors, including historical stock price data and sentiment contained in financial news. This research aims to develop a more accurate stock price prediction model by utilizing Spatial-Temporal Attention-Based Convolutional Network (STACN). This model is designed to predict stock prices using historical stock data and information from financial news. The methodology involves integrating Convolutional Neural Network (CNN) to extract features from news thought vectors, Long Short-Term Memory (LSTM) to capture temporal patterns from stock price data, and Spatial-Temporal Attention Network (STAN) to assign attention weights to relevant features. A case study was conducted on energy sector stocks listed on the Indonesia Stock Exchange (IDX), using historical stock price data and news from Indonesian business portals. This research examines the effectiveness of the STACN model with various architectural configurations, including full bidirectional (full_bi), full convolutional (full_conv), no attention convolutional, and pure LSTM-based models. Model performance evaluation was conducted using metrics such as MAE (Mean Absolute Error), RMSE (Root Mean Square Error), MAPE (Mean Absolute Percentage Error), and R² (R-squared). Research results show that the STACN model with full bidirectional configuration consistently provides the best performance, with the lowest MAPE values ranging from 0.58% to 1.61%. These findings indicate the importance of the bidirectional component in capturing complex temporal patterns and the integration of text and numerical data to improve prediction accuracy. The results of this research can serve as a reference for the development of stock price prediction models based on historical data and financial news, particularly in understanding how semantic representations of financial news can influence market movements.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > QA Mathematics > QA76.6 Computer Programming | ||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | Novita Anggraini | ||||||||||||
| Date Deposited: | 17 Mar 2025 04:47 | ||||||||||||
| Last Modified: | 17 Mar 2025 04:47 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/35569 |
Actions (login required)
 |
View Item |