Analisis Prediksi Harga dan Volatilitas Saham IDX30 Dengan Pendekatan Support Vector Regression dan XGBoost Reggressor

Prasetyo, Jessyca Frista Herdiana (2026) Analisis Prediksi Harga dan Volatilitas Saham IDX30 Dengan Pendekatan Support Vector Regression dan XGBoost Reggressor. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221083-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (16MB) | Request a copy

Abstract

Pasar saham merupakan salah satu instrumen investasi yang memiliki potensi keuntungan tinggi, namun juga diiringi oleh tingkat ketidakpastian dan risiko yang besar. Oleh karena itu, prediksi harga penutupan saham dan volatilitas log return menjadi penting untuk mendukung pengambilan keputusan investasi. Penelitian ini bertujuan membandingkan kinerja metode Support Vector Regression (SVR) dan XGBoost Regressor (XGBR) dalam memprediksi harga penutupan saham dan volatilitas log return saham IDX30 berdasarkan metrik Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), dan Coefficient of Determination (R²). Data yang digunakan berupa harga penutupan harian 28 saham IDX30 periode 15 Januari 2018 hingga 15 Januari 2026 yang diperoleh dari Yahoo Finance. Tahapan penelitian meliputi feature engineering menggunakan variabel input berupa lag price, lag log return, volatilitas historis, Simple Moving Average (SMA), dan Exponential Moving Average (EMA), dilanjutkan dengan preprocessing data, pembagian data latih dan data uji, penentuan hyperparameter terbaik menggunakan Time Series Cross Validation (TimeSeriesSplit), pemodelan, serta evaluasi kinerja model. Hasil penelitian menunjukkan bahwa kedua metode mampu digunakan untuk memprediksi harga penutupan saham maupun volatilitas log return. Kedua model memperoleh performa terbaik pada horizon prediksi satu hari (t+1), sedangkan akurasi cenderung menurun seiring bertambahnya horizon prediksi. Berdasarkan hasil evaluasi menggunakan metrik MAE, RMSE, MAPE, dan R², Support Vector Regression (SVR) secara umum memberikan performa yang lebih baik dibandingkan XGBoost Regressor, baik pada prediksi harga penutupan saham maupun volatilitas log return. Hasil pemeringkatan juga menunjukkan bahwa KLBF merupakan saham dengan performa prediksi terbaik secara konsisten, sedangkan SMGR merupakan saham yang paling sulit diprediksi. Hasil penelitian menunjukkan bahwa tingkat akurasi prediksi dipengaruhi oleh metode yang digunakan, horizon prediksi, serta karakteristik historis masing-masing saham.
====================================================================================================================================
The stock market is one of the investment instruments that offers high return potential but is accompanied by considerable uncertainty and risk. Therefore, predicting stock closing prices and log return volatility is essential to support investment decision-making. This study aims to compare the performance of Support Vector Regression (SVR) and XGBoost Regressor (XGBR) in predicting the closing prices and log return volatility of IDX30 stocks using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R²) as evaluation metrics. The dataset consists of daily closing prices of 28 IDX30 stocks from January 15, 2018, to January 15, 2026, obtained from Yahoo Finance. The research procedure includes feature engineering using lag prices, lag log returns, historical volatility, Simple Moving Average (SMA), and Exponential Moving Average (EMA), followed by data preprocessing, training-testing data splitting, hyperparameter tuning using Time Series Cross Validation (TimeSeriesSplit), model development, and performance evaluation. The results indicate that both SVR and XGBR are capable of predicting stock closing prices and log return volatility. The best prediction performance for both models was generally achieved at the one-day prediction horizon (t+1), while prediction accuracy decreased as the prediction horizon increased. In predicting stock closing prices, SVR outperformed XGBR, producing lower prediction errors and higher R² values for the majority of stocks. Similarly, for log return volatility prediction, SVR generally achieved better performance, although XGBR produced superior results for several individual stocks. Furthermore, the ranking analysis showed that KLBF consistently had the best prediction performance, whereas SMGR was the most difficult stock to predict. Overall, the findings suggest that prediction performance is influenced not only by the forecasting method but also by the historical characteristics of each stock.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Harga Saham, IDX30, Support Vector Regression, Volatilitas, XGBoost Regressor.
Subjects:	Q Science
Divisions:	Faculty of Mathematics, Computation, and Data Science > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Jessyca Frista Herdiana Prasetyo
Date Deposited:	20 Jul 2026 03:39
Last Modified:	20 Jul 2026 03:39
URI:	http://repository.its.ac.id/id/eprint/135629

Actions (login required)

View Item