Implementasi Temporal Feature Extraction Hidden Markov Model (TFE-HMM) Dalam Prediksi Arah Pergerakan Harga Saham Di Bursa Efek Indonesia

Ramadhan, Rafi Adyatma (2026) Implementasi Temporal Feature Extraction Hidden Markov Model (TFE-HMM) Dalam Prediksi Arah Pergerakan Harga Saham Di Bursa Efek Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221045-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Penelitian ini mengimplementasikan metode Temporal Feature Extraction Hidden Markov Model (TFE-HMM) untuk memprediksi arah pergerakan harga saham pada tiga papan pencatatan Bursa Efek Indonesia menggunakan data harga harian periode Januari 2023 - Juli 2025. Model ini menggabungkan hidden state hasil Gaussian HMM dengan tren harga dalam sliding window, kemudian mengukur kesamaan pola melalui weighted mixed similarity untuk menghasilkan prediksi arah harga berikutnya. Hasil pengujian menunjukkan bahwa performa model sangat dipengaruhi pemilihan hyperparameter, dengan akurasi terbaik 50,88% (KLBF), 52,98% (PANI), dan 47,00% (KLIN) yang menunjukkan kemampuan prediksi sedikit di atas acak pada dua dataset paling stabil. Simulasi strategi perdagangan harian juga memperlihatkan bahwa strategi long-short jauh lebih representatif dibanding buy-hold, menghasilkan return kumulatif tertinggi 0,25999, 0,26710, dan 0,26768 pada masing-masing saham. Selain itu, model menunjukkan sensitivitas yang kuat terhadap volatilitas dan struktur pola temporal tiap emiten, sehingga konfigurasi optimal tiap dataset berbeda secara signifikan. Secara keseluruhan, penelitian ini menegaskan bahwa TFE-HMM mampu menangkap dinamika harga secara adaptif meskipun stabilitas prediksi sangat bergantung pada karakteristik data, serta memberikan kontribusi metodologis dan referensi praktis bagi investor dalam mengevaluasi strategi perdagangan berbasis arah harga.
======================================================================================================================================
This study implements the Temporal Feature Extraction Hidden Markov Model (TFE-HMM) to predict stock price movement direction across three listing boards of the Indonesia Stock Exchange using daily price data from January 2023 to July 2025. The model integrates hidden states generated by a Gaussian HMM with price trends within a sliding temporal window, and measures pattern similarity using a weighted mixed similarity framework to produce next-day directional predictions. The results show that model performance is highly dependent on hyperparameter selection, with the best accuracies of 50.88% (KLBF), 52.98% (PANI), and 47.00% (KLIN), indicating that the predictive ability is only slightly above random for the two more stable datasets. Daily trading simulations further demonstrate that the long-short strategy is significantly more representative than buy-hold, producing the highest cumulative returns of 0.25999, 0.26710, and 0.26768 respectively. In addition, the model exhibits strong sensitivity to volatility levels and temporal structure variations across stocks, leading to distinct optimal configurations for each dataset. Overall, the findings indicate that TFE-HMM can adaptively capture complex market dynamics despite its moderate stability, while providing both methodological contributions and practical insights for investors evaluating direction-based trading strategies.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bursa Efek indonesia, Prediksi Arah Pergerakan Harga Saham, Simulasi Perdagangan Harian, TFE-HMM, Weighted Mixed Similarity, Daily Trading Simulation, Indonesia Stock Exchange, Stock Price Movement Prediction, TFE-HMM, Weighted Mixed Similarity
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HG Finance > HG4529 Investment analysis H Social Sciences > HG Finance > HG4915 Stocks--Prices Q Science > QA Mathematics > QA274.7 Markov processes--Mathematical models.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Rafi Adyatma Ramadhan
Date Deposited:	14 Jan 2026 07:08
Last Modified:	14 Jan 2026 07:08
URI:	http://repository.its.ac.id/id/eprint/129612

Actions (login required)

View Item