Prediksi Harga Saham Intraday Menggunakan Artificial Intelligence di ASEAN-5

Ria Anastasia, Dr. Leo Indra Wardhana, M. Sc., CFP., GRCE.

2026 | Tesis | S2 Magister Ek.Pembangunan

Penelitian ini mengevaluasi kinerja Artificial Intelligence untuk prediksi harga saham intraday pada pasar saham ASEAN-5 menggunakan model Long Short-Term Memory (LSTM) dengan dua skenario input. Skenario pertama memakai data harga historis, sedangkan skenario kedua mengombinasikan harga historis dengan indikator teknikal Simple Moving Average (periode 5, 14, dan 50 hari) serta HLC3. Akurasi prediksi model dievaluasi menggunakan MAPE dan MAE. Hasil menunjukkan error prediksi relatif rendah di seluruh bursa, tetapi tidak homogen antar target harga intraday. Pola yang paling konsisten adalah MAPE terendah pada prediksi Open Price, meningkat pada High dan Low Price, serta tertinggi pada prediksi Close Price. pada skenario input hanya historis, MAPE minimum 0,11?n maksimum 3,55%. pada skenario combined input, minimum 0,17?n maksimum 5,51%. Perbedaan kinerja lintas negara terutama dipicu outlier pada sebagian emiten. Penambahan indikator teknikal tidak konsisten meningkatkan akurasi, sehingga manfaatnya bersifat kontekstual dan dipengaruhi karakteristik emiten serta dinamika pasar.

This study evaluates the performance of Artificial Intelligence for intraday stock Price forecasting in ASEAN-5 equity markets using a Long Short-Term Memory (LSTM) model under two input scenarios. The first scenario uses historical Price data only, while the second combines historical prices with technical indicators, namely Simple Moving Average (5, 14, and 50 day) and HLC3. Forecast accuracy is assessed using MAPE and MAE. The results indicate relatively Low forecasting errors across all exchanges, yet performance is not homogeneous across intraday Price targets. The most consistent pattern is the Lowest MAPE for Open Price forecasts, Higher errors for High and Low Price forecasts, and the Highest errors for Close Price forecasts. Under the historical-only scenario, MAPE ranges from 0.11% to 3.55%. Under the combined-input scenario, MAPE ranges from 0.17% to 5.51%. Cross-country differences are largely driven by outliers in a subset of stocks. Adding technical indicators does not consistently improve accuracy, implying that their usefulness is context-dependent and shaped by firm characteristics and market dynamics.

Kata Kunci : LSTM, Prediksi harga saham, harga historis saham, Indikator teknikal, ASEAN-5