PARAMANANDA, Aditya and Warsito, Budi and Surarso, Bayu (2026) MODEL PREDIKSI HARGA SAHAM BLUE CHIP MENGGUNAKAN MULTILAYER PERCEPTRON DENGAN PENDEKATAN SELEKSI FITUR FORWARD SELECTION DAN TUNING HYPERPARAMETER GRID SEARCH. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text cover aja.pdf Download (141kB)
	Text Cover lengkap.pdf Restricted to Repository staff only Download (1MB)
	Text BAB I.pdf Download (236kB)
	Text BAB II.pdf Download (540kB)
	Text BAB III.pdf Restricted to Repository staff only Download (527kB)
	Text BAB IV.pdf Restricted to Repository staff only Download (2MB)
	Text BAB V.pdf Restricted to Repository staff only Download (295kB)
	Text Daftar Pustaka.pdf Download (245kB)
	Text Lampiran.pdf Restricted to Repository staff only Download (241kB)

Abstract

Ketidakpastian geopolitik dan ekonomi global telah meningkatkan volatilitas pasar keuangan global, termasuk pasar modal Indonesia. Kondisi tersebut mendorong investor menerapkan strategi flight to quality dengan mengalihkan investasinya ke saham blue chip. Kemampuan memprediksi pergerakan harga saham blue chip menjadi semakin penting untuk mendukung pengambilan keputusan investasi, pengelolaan risiko, dan penyusunan strategi investasi. Perkembangan Artificial Neural Network (ANN) khususnya Multilayer Perceptron (MLP), menawarkan pendekatan yang lebih adaptif dalam menangkap pola nonlinier pada data deret waktu. Kinerja MLP sangat dipengaruhi oleh pemilihan fitur yang relevan dan konfigurasi hyperparameter yang optimal. Penelitian ini mengusulkan integrasi metode seleksi fitur Forward Selection dan tuning hyperparameter Grid Search pada model MLP untuk meningkatkan kinerja prediksi harga saham. Penelitian ini menggunakan data historis harga saham PT Bank Central Asia (BCA) dan PT Astra sebagai objek penelitian. Hasil penelitian menunjukkan bahwa model MLP mampu memprediksi harga saham dengan baik. Integrasi model MLP dengan Forward Selection dan Grid Search, menghasilkan R² sebesar 0,8901, MAE sebesar 0,0289, dan RMSE sebesar 0,0418 untuk harga saham BCA, dan pada harga saham Astra menghasilkan R² sebesar 0,9815, MAE 0,0262, dan RMSE 0,0365. Hasil ini membuktikan bahwa integrasi seleksi fitur dan optimasi hyperparameter MLP mampu menghasilkan prediksi yang lebih baik dibandingkan model lainnya.
Kata Kunci: Prediksi Harga Saham, Saham Blue Chip, Multi-Layer Perceptron, Seleksi Fitur Forward Selection, Optimasi Hiperparameter Grid Search

Global geopolitical and economic uncertainty has increased the volatility of global financial markets, including the Indonesian capital market. This condition encourages investors to implement a "flight to quality" strategy by shifting their investments to blue chip stocks. The ability to predict blue chip stock prices is increasingly important to support investment decision making, risk management and investment strategy development. The development of Artificial Neural Networks (ANNs), particularly Multilayer Perceptron (MLPs), offers a more adaptive approach to capturing nonlinear patterns in time-series data. MLP performance is strongly influenced by the selection of relevant features and optimal hyperparameter configuration. This study proposes the integration of Forward Selection feature selection methods and Grid Search hyperparameter tuning into the MLP model to improve stock price prediction performance. This study uses historical stock price data from PT Bank Central Asia (BCA) and PT Astra as the research objects. The results show that the MLP model is capable of predicting stock prices well. The integration of the MLP model with Forward Selection and Grid Search, produces an R² of 0.8901, MAE of 0.0289, and RMSE of 0.0418 for BCA stock prices, and for Astra stock prices produces an R² of 0.9815, MAE of 0.0262, and RMSE of 0.0365. These results prove that the integration of feature selection and hyperparameter optimization of MLP is able to produce better predictions than other models.
Keywords: Stock Price Forecasting, Blue-Chip Stocks, Multi-Layer Perceptron, Forward Selection, Grid Search Hyperparameter Optimization

Actions (login required)

View Item