Laporkan Masalah

Perbandingan Particle Swarm Optimization, Genetic Algorithm, dan Artificial Bee Colony pada Optimasi Portofolio Saham Jakarta Islamic Index Berbasis Klasterisasi K-Means++ dan Expected Shortfall

Hayatul Husnah Sopiani, Prof. Dr. Drs. Gunardi, M.Si.

2026 | Skripsi | STATISTIKA

Pertumbuhan jumlah investor pasar modal Indonesia sekitar 35% pada periode 2024–2025 menunjukkan meningkatnya minat terhadap investasi saham, meskipun disertai risiko tinggi akibat fluktuasi harga. Model mean-variance memiliki keterbatasan dalam merepresentasikan tail risk karena mengasumsikan distribusi return normal dan mengukur risiko secara simetris. Penelitian ini bertujuan mengelompokkan saham Jakarta Islamic Index (JII) menggunakan K-Means++, membentuk portofolio optimal dengan Particle Swarm Optimization (PSO), Genetic Algorithm (GA), dan Artificial Bee Colony (ABC) berdasarkan Expected Shortfall (ES), serta membandingkan kinerja ketiganya. Data yang digunakan berupa harga penutupan harian 30 saham konstituen JII dari Yahoo Finance yang diolah melalui klasterisasi K-Means++ dan model Mean-Expected Shortfall. Hasil penelitian menunjukkan terbentuknya empat klaster optimal dengan saham representatif INCO.JK, INDF.JK, ADRO.JK, dan BRPT.JK. Ketiga algoritma menghasilkan Sharpe Ratio sekitar 15% lebih tinggi dibandingkan portofolio equal weight. Berdasarkan nilai kinerja, analisis sensitivitas, dan kesesuaian dengan ES, Artificial Bee Colony memberikan hasil paling konsisten. Temuan ini menunjukkan bahwa integrasi K-Means++ dan Expected Shortfall efektif meningkatkan kualitas portofolio saham syariah serta memperkuat bukti empiris mengenai penerapan ABC. Penelitian selanjutnya disarankan mempertimbangkan biaya transaksi, melakukan pengujian out-of-sample, dan memperluas objek penelitian pada indeks saham syariah lainnya.

The approximately 35% increase in the number of Indonesian capital market investors during the 2024–2025 period reflects growing interest in stock investment despite the high risk associated with price fluctuations. The mean-variance model has limitations in capturing tail risk because it assumes normally distributed returns and measures risk symmetrically. This study aims to cluster stocks listed in the Jakarta Islamic Index (JII) using K-Means++, construct optimal portfolios using Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Artificial Bee Colony (ABC) based on Expected Shortfall (ES), and compare their performance. The study uses daily closing prices of 30 JII constituent stocks obtained from Yahoo Finance and applies K-Means++ clustering followed by the Mean-Expected Shortfall optimization model. The results identify four optimal clusters with INCO.JK, INDF.JK, ADRO.JK, and BRPT.JK as representative stocks. All three algorithms produce portfolios with Sharpe Ratios approximately 15% higher than the equal-weight portfolio. Based on portfolio performance, sensitivity analysis, and consistency with the ES risk measure, the Artificial Bee Colony algorithm delivers the most consistent results. These findings demonstrate that integrating K-Means++ and Expected Shortfall effectively improves Islamic stock portfolio quality while providing additional empirical evidence for the application of the ABC algorithm. Future studies are encouraged to incorporate transaction costs, conduct out-of-sample testing, and extend the analysis to other Islamic stock indices.

Kata Kunci : Jakarta Islamic Index, K-Means++, Expected Shortfall, optimasi portofolio, Artificial Bee Colony

  1. S1-2026-497083-abstract.pdf  
  2. S1-2026-497083-bibliography.pdf  
  3. S1-2026-497083-tableofcontent.pdf  
  4. S1-2026-497083-title.pdf