Stock Market Time Series Forecasting: Bridging Traditional Machine Learning And Deep Learning Approaches

Ardiansyah, Abdul Aziz (2026) Stock Market Time Series Forecasting: Bridging Traditional Machine Learning And Deep Learning Approaches. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026201080-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Prediksi pasar saham masih menjadi area studi yang utama dan sangat menarik karena dampaknya terhadap pengambilan keputusan keuangan dan manajemen risiko. Metode statistik konvensional mengalami kesulitan dalam memperhitungkan non-linearitas dan fluktuasi harga yang dinamis di pasar keuangan yang semakin kompleks. Berdasarkan data historis dari tahun 2010 hingga 2020, penelitian ini menilai seberapa baik berbagai model machine learning dan deep learning dalam memprediksi pola saham pada tiga perusahaan multinasional yang berbeda, yaitu Samsung yang merepresentasikan Asia, SAP yangmerepresentasikan Eropa, dan Apple Inc. yang merepresentasikan Amerika Serikat. Support Vector Regression (SVR), K Nearest Neighbors (KNN), Random Forest, dan XGBoost merupakan algoritma machine learning klasik yang dibandingkan dalam penelitian ini dengan model deep learning hibrida yang canggih, yang mengombinasikan Long Short-Term Memory (LSTM) dan Convolutional Neural Network (CNN). Kinerja model diukur menggunakan Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), dan Directional Accuracy.Hasil penelitian menunjukkan bahwa model hibrida LSTM+CNN secara konsisten mengungguli model-model klasik dalam akurasi prediksi harga (dengan nilai MSE, MAE, dan MAPE terendah), sementara akurasi arah pergerakan harga (directional accuracy) tetap menjadi tantangan umum bagi seluruh model.
====================================================================================================================================
Predict stock market that is still a major and really interesting study area due to its impact on financial decision making and risk management. Conventional statistical methods find it difficult to account for nonlinearities and dynamic price fluctuations in the increasingly complex financial markets. Based on historical data from 2010 to 2020, this study assesses how well different machine learning and deep learning models predict stock patterns in 3 different multinational corporations, which is: Samsung that represent Asia, SAP represent Europe, and Apple Inc. that reprsent US. Support Vector Regression (SVR), K Nearest Neighbors (KNN), Random Forest, and XGBoost are classic machine learning algorithms’ that are compared in this study to a sophisticated hybrid deep learning model that combines ‘Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN). Model performances are measured using Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Directional Accuracy. Results demonstrate that the LSTM+CNN hybrid consistently outperforms classical models in price prediction accuracy (lowestMSE, MAE, and MAPE), while directional accuracy remains a common challenge for all models.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Peramalan Pasar Saham, Pembelajaran Mesin, Pembelajaran Mendalam, LSTM, CNN, Analisis Deret Waktu, Dasbor Web, Random Forest, XGBoost, SVR. ============================================================ Stock Market Forecasting, Machine Learning, Deep Learning, LSTM, CNN, Time-Series Analysis, Web Dashboard, Random Forest, XGBoost, SVR.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Abdul Aziz Ardiansyah
Date Deposited:	31 Jan 2026 06:49
Last Modified:	31 Jan 2026 06:49
URI:	http://repository.its.ac.id/id/eprint/131423

Actions (login required)

View Item