Prediksi Harga Saham Menggunakan Pendekatan Deep Learning Dengan Metode Convolutional Neural Network dan Long Short-Term Memory

Hidiya, Fais Rafii Akbar (2023) Prediksi Harga Saham Menggunakan Pendekatan Deep Learning Dengan Metode Convolutional Neural Network dan Long Short-Term Memory. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111940000026-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (11MB) | Request a copy

Abstract

Perubahan harga saham diidentifikasi sebagai isu kritis dalam bidang ekonomi, dipengaruhi oleh beragam faktor internal dan eksternal. Metode analisis tradisional seperti analisis fundamental dan teknikal, serta model time series seperti VAR, BVAR, ARIMA, dan GARCH, telah banyak digunakan, namun akurasinya masih dipertanyakan. Oleh karena itu, penelitian ini mengusulkan solusi berupa penggunaan metode gabungan Convolutional Neural Network (CNN) dan variasi Long Short-Term Memory (LSTM) untuk meningkatkan akurasi prediksi harga saham. Metode ini memanfaatkan CNN untuk mengekstraksi fitur dari input time data dan menggunakan Bidirectional LSTM untuk memprediksi harga penutupan saham di hari berikutnya. Pengumpulan data melibatkan pengumpulan data harga dari 10 saham di Bursa Efek Indonesia dari periode 1 November 2018 hingga 1 November 2022. Proses pemilihan fitur dilakukan untuk memastikan dataset dapat digunakan, dan beberapa atribut dipilih termasuk harga penutupan, dan pergerakan rata-rata harga saham dalam 10, 20, 50, dan 100 hari terakhir. Setelah pra-pemrosesan data dan pemisahan data menjadi data pelatihan dan pengujian, model CNN-Bi-LSTM dilatih dengan data ini. Proses ini juga melibatkan tuning parameter untuk meningkatkan akurasi model. Pengujian dan evaluasi model menunjukkan peningkatan signifikan dalam akurasi prediksi. Skenario terbaik adalah menggunakan Keras Tuner dalam pemilihan hyperparameter, menghasilkan Mean Absolute Percentage Error (MAPE) terendah 1,31 dan Root Mean Squared Error (RMSE) terendah 92,26. Dengan demikian, kombinasi CNN dan LSTM dengan penyesuaian hyperparameter yang tepat dapat memberikan performa yang lebih baik dalam memprediksi harga saham. Penelitian ini menunjukkan bahwa penggunaan teknologi deep learning seperti CNN dan variasi LSTM dapat memberikan wawasan baru dan peningkatan akurasi dalam prediksi harga saham.
=================================================================================================================================
Stock price fluctuation is identified as a critical issue in the field of economics, influenced by various internal and external factors. Traditional analysis methods such as fundamental and technical analysis, as well as time-series models like VAR, BVAR, ARIMA, and GARCH, have been widely used, but their accuracy remains questionable. Therefore, this study proposes a solution in the form of a combined method of Convolutional Neural Network (CNN) and a variant of Long Short-Term Memory (LSTM) to improve the accuracy of stock price prediction. This method leverages CNN to extract features from time series data inputs and uses Bidirectional LSTM to predict the closing stock price of the next day. The data collection involves gathering the price data of 10 stocks on the Indonesia Stock Exchange from the period of November 1, 2018, to November 1, 2022. Feature selection processes are conducted to ensure that the dataset can be utilized, and several attributes are chosen, including closing prices and average stock price movements over the last 10, 20, 50, and 100 days. After pre-processing the data and separating it into training and testing data, the CNN-Bi-LSTM model is trained with this data. This process also involves tuning parameters to enhance the model's accuracy. Testing and evaluation of the model show significant improvement in prediction accuracy. The best scenario involves using Keras Tuner for hyperparameter selection, yielding the lowest Mean Absolute Percentage Error (MAPE) of 1.31 and the lowest Root Mean Squared Error (RMSE) of 92.26. Thus, the combination of CNN and LSTM with appropriate hyperparameter tuning can provide superior performance in predicting stock prices. This research suggests that the utilization of deep learning technologies such as CNN and LSTM variants can offer new insights and improved accuracy in stock price predictions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deep Learning, Prediksi Harga Saham, Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), Time-Series Analysis
Subjects:	T Technology > T Technology (General) 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:	Fais Rafii Akbar Hidiya
Date Deposited:	09 Oct 2023 02:14
Last Modified:	09 Oct 2023 02:14
URI:	http://repository.its.ac.id/id/eprint/102828

Actions (login required)

View Item