Optimasi Portofolio Saham IDX30 Menggunakan Black-Litterman dengan View Berbasis Bidirectional Long Short-Term Memory (Bi-LSTM) dan Hierarchical Risk Parity (HRP)

Anggarjati, Marva Rizqi (2026) Optimasi Portofolio Saham IDX30 Menggunakan Black-Litterman dengan View Berbasis Bidirectional Long Short-Term Memory (Bi-LSTM) dan Hierarchical Risk Parity (HRP). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221096-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Optimasi portofolio pada indeks IDX30 menghadapi tantangan berupa risiko konsentrasi dan korelasi tinggi antar saham berkapitalisasi besar, yang membatasi efektivitas diversifikasi konvensional berbasis Mean-Variance Optimization (MVO). Penelitian ini mengintegrasikan model prediksi Bidirectional Long Short-Term Memory (Bi-LSTM) ke dalam kerangka Black-Litterman untuk menghasilkan expected return posterior yang lebih adaptif, kemudian menerapkan metode Hierarchical Risk Parity (HRP) berbasis matriks kovarians posterior untuk alokasi bobot portofolio. Data yang digunakan adalah harga penutupan harian 21 saham IDX30 selama periode Januari 2019 hingga Februari 2026. Model Bi-LSTM dilatih untuk memprediksi harga saham (dalam skala Z-Score), yang kemudian dikonversi menjadi return harian sebagai dasar pembentukan pandangan investor (views) dalam model Black-Litterman. Kepercayaan pandangan dikalibrasi menggunakan Mean Squared Error (MSE) prediksi Bi-LSTM melalui matriks ketidakpastian pandangan investor (Ω). Hasil penelitian menunjukkan bahwa model Bi-LSTM mampu memprediksi return harian 21 saham IDX30 dengan rata-rata MSE (Mean Squared Error) sebesar 0,0017 pada periode testing 175 hari. Portofolio BL-HRP yang menggunakan kovarians posterior Black-Litterman menunjukkan kinerja lebih baik dibandingkan HRP baseline pada tiga dari empat metrik evaluasi pada periode out-of-sample. Secara rinci, BL-HRP mencatatkan Cumulative Return (CR) sebesar 13,18%, Sharpe Ratio sebesar 0,8966, Maximum Drawdown sebesar −6,97%, dan Calmar Ratio sebesar 2,7414; sedangkan HRP baseline secara berturut-turut mencapai 11,03%, 0,7327, −6,92%, dan 2,3539. Meskipun HRP baseline memiliki Maximum Drawdown sedikit lebih rendah, BL-HRP tetap menunjukkan kinerja keseluruhan yang lebih baik dari sisi pertumbuhan nilai portofolio, kinerja berbasis risiko, dan rasio return terhadap drawdown. Selain portofolio utama, penelitian juga membentuk empat mini-portofolio berbasis segmentasi klaster HRP, yang dikarakterisasi menggunakan indikator fundamental Return on Equity (ROE). Temuan ini menunjukkan bahwa integrasi Bi-LSTM, Black-Litterman, dan HRP menghasilkan kerangka optimasi portofolio yang lebih efisien dan adaptif dibandingkan pendekatan berbasis kovarians historis.
=======================================================================================================================================
Portfolio optimization of the IDX30 index faces challenges arising from concentration risk and high correlation among large-capitalization stocks, which limits the effectiveness of conventional diversification approaches such as Mean-Variance Optimization (MVO). This study integrates a Bidirectional Long Short-Term Memory (Bi-LSTM) prediction model into the Black-Litterman framework to generate adaptive posterior expected returns, and subsequently applies Hierarchical Risk Parity (HRP) based on the posterior covariance matrix for portfolio weight allocation. The data used consist of daily closing prices of 21 IDX30 constituent stocks over the period January 2019 to February 2026. The Bi-LSTM model was trained to predict stock prices (in Z-Score scale), which were subsequently converted into daily returns as the basis for forming investor views within the Black-Litterman model, with view confidence calibrated through the uncertainty matrix (Ω) derived from Bi-LSTM prediction MSE. The results show that the Bi-LSTM model achieved an average MSE of 0.0017 across 21 stocks over a 175-day testing period. The BL-HRP portfolio, which incorporates the Black-Litterman posterior covariance matrix, outperformed the HRP baseline in three out of four evaluation metrics during the out-of-sample period. Specifically, BL-HRP recorded a Cumulative Return (CR) of 13.18%, a Sharpe Ratio of 0.8966, a Maximum Drawdown of −6.97%, and a Calmar Ratio of 2.7414, whereas the HRP baseline achieved 11.03%, 0.7327, −6.92%, and 2.3539, respectively. Although the HRP baseline produced a slightly lower Maximum Drawdown, BL-HRP demonstrated better overall performance in terms of portfolio growth, risk-adjusted return, and return relative to drawdown risk. In addition to the main portfolio, the study constructs four mini-portfolios based on HRP cluster segmentation, characterized using the Return on Equity (ROE) fundamental indicator. These findings demonstrate that the integration of Bi-LSTM, Black-Litterman, and HRP produces a more efficient and adaptive portfolio optimization framework compared to approaches based on historical covariance.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bi-LSTM, Black-Litterman, Hierarchical Risk Parity, IDX30, Optimasi Portofolio, Portfolio Optimization
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HG Finance > HG4012 Mathematical models H Social Sciences > HG Finance > HG4529 Investment analysis H Social Sciences > HG Finance > HG4529.5 Portfolio management H Social Sciences > HG Finance > HG4910 Investments H Social Sciences > HG Finance > HG4915 Stocks--Prices
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Marva Rizqi Anggarjati
Date Deposited:	17 Jul 2026 04:31
Last Modified:	17 Jul 2026 04:31
URI:	http://repository.its.ac.id/id/eprint/135167

Actions (login required)

View Item