Perbandingan Kinerja Model Multivariate LSTM dalam Memprediksi Harga Saham Tambang di Indonesia
Date
2026Author
ALDENA, MAULANA TATA
Budiarti, Retno
Ardana, Ngakan Komang Kutha
Metadata
Show full item recordAbstract
Fluktuasi harga saham sektor pertambangan periode 2020-2025 menjadi tantangan dalam prediksi standar. Penelitian ini mengevaluasi kinerja model multivariate Long Short-Term Memory (LSTM) dalam memprediksi harga harian saham ADRO, ANTM, dan MDKA. Hasil menunjukkan model LSTM mampu menjelaskan lebih dari 89% varians data pada ADRO dan MDKA, dengan MDKA mendemonstrasikan performa paling optimal dan konsisten (??^2 > 0.93, MSE < 0.00041). Meskipun volatilitas tinggi pada saham ANTM memengaruhi akurasi,
model tetap mampu memberikan hasil memadai pada harga penutupan dengan nilai ??^2 sebesar 0.85681. Uji Friedman mengonfirmasi adanya perbedaan performa prediksi antar saham pada variabel harga pembukaan, sementara variabel harga penutupan cenderung stabil. Secara keseluruhan, model LSTM pada data saham
MDKA terbukti sebagai metode peramalan yang paling konsisten dibandingkan sampel lainnya. The fluctuation of stock prices in the mining sector from 2020-2025 poses a challenge for standard prediction methods. This study evaluates the performance of the multivariate Long Short-Term Memory (LSTM) model in predicting daily
stock prices for ADRO, ANTM, and MDKA. Results show that the LSTM model is capable of explaining more than 89% of the data variance for ADRO and MDKA, with MDKA demonstrating the most optimal and consistent performance (??^2 >
0.93, MSE < 0.00041). Although high volatility in ANTM stock affects accuracy, the model remains capable of providing adequate results for closing prices with an ??^2 value of 0.85681. Friedman tests confirm the existence of differences in
prediction performance among stocks for the opening price variable, while the closing price variable tends to be stable. Overall, the LSTM model on MDKA stock data is proven to be the most consistent forecasting method compared to the other samples.
Collections
- UF - Actuaria [99]

