Optimasi Portofolio Komoditas Global Lintas Sektor Menggunakan Genetic Algorithm dengan Fungsi Fitness Mean-Variance Markowitz

Sari, Novita Indah (2026) Optimasi Portofolio Komoditas Global Lintas Sektor Menggunakan Genetic Algorithm dengan Fungsi Fitness Mean-Variance Markowitz. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221122-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Volatilitas harga komoditas global yang tinggi menyebabkan pembentukan portofolio menjadi semakin kompleks sehingga diperlukan metode optimasi yang mampu menghasilkan keseimbangan antara return dan risiko. Penelitian ini bertujuan untuk mengimplementasikan Genetic Algorithm (GA) dengan fungsi fitness Mean-Variance Markowitz yang terintegrasi dengan cardinality constraint dan diversification constraint untuk menyusun portofolio optimal komoditas global lintas sektor energi, logam, dan pertanian. Data yang digunakan berupa harga penutupan harian 20 komoditas global pada tahun 2015−2026 yang diperoleh dari Yahoo Finance. Penelitian dilakukan dengan pengujian 27 kombinasi parameter meliputi ukuran populasi, pc, dan pm dengan setiap konfigurasi diuji pada nilai λ dari 0,5 hingga 5,0. Konfigurasi parameter terbaik diperoleh pada ukuran populasi 100, generasi 300, pc 0,8, pm 0,1 dengan λ = 1,5. Komposisi portofolio optimal terdiri dari 7 aset, yaitu Gold 32,5%, Crude Oil WTI 20%, Rough Rice 18,8%, Iron Ore 12,4%, Cocoa 8,2%, Sugar 5,8%, dan Soybean 2,3%. Pada data latih, portofolio menghasilkan return tahunan sebesar 6,4833%, risiko 14,0956%, dan sharpe ratio 0,4600, sedangkan pada data uji meningkat menjadi return 7,8370%, risiko 13,9558%, dan sharpe ratio 0,5616. Kualitas solusi yang dihasilkan GA kemudian dievaluasi menggunakan solusi eksak Mixed Integer Quadratic Programming (MIQP) sebagai acuan. Hasil evaluasi menunjukkan bahwa GA menghasilkan nilai fitness sebesar 0,03503 yang sangat mendekati solusi MIQP sebesar 0,03510, meskipun memerlukan waktu komputasi yang lebih lama, yaitu 9,7 detik dibandingkan 1,51 detik pada MIQP. Hasil tersebut menunjukkan bahwa GA mampu menghasilkan solusi yang mendekati optimum sehingga layak digunakan untuk optimasi portofolio komoditas lintas sektor.
======================================================================================================================================
The high price volatility of global commodities makes portfolio construction increasingly complex, requiring an optimization method capable of achieving a balance between return and risk. This study aims to implement a Genetic Algorithm (GA) using the Mean-Variance Markowitz fitness function integrated with cardinality and diversification constraints to construct an optimal global commodity portfolio across the energy, metals, and agricultural sectors. The dataset consists of the daily closing prices of 20 global commodities from 2015 to 2026, obtained from Yahoo Finance. The proposed approach was evaluated using 27 parameter combinations involving population size, crossover probability (pc), and mutation probability (pm), with each configuration tested across λ values ranging from 0.5 to 5.0. The best parameter configuration was obtained with a population size of 100, 300 generations, pc = 0.8, pm = 0.1, and λ = 1.5. The resulting optimal portfolio consists of seven assets: Gold (32.5%), Crude Oil WTI (20.0%), Rough Rice (18.8%), Iron Ore (12.4%), Cocoa (8.2%), Sugar (5.8%), and Soybean (2.3%). On the training dataset, the portfolio achieved an annual return of 6.4833%, a risk of 14.0956%, and a Sharpe ratio of 0.4600. On the testing dataset, the portfolio performance improved, yielding an annual return of 7.8370%, a risk of 13.9558%, and a Sharpe ratio of 0.5616. The quality of the solution produced by the GA was subsequently evaluated using the exact solution obtained from Mixed Integer Quadratic Programming (MIQP) as a benchmark. The evaluation results indicate that the GA achieved a fitness value of 0.03503, which is very close to the MIQP solution of 0.03510, although it required a longer computation time (9.7 seconds compared to 1.51 seconds for MIQP). These findings demonstrate that the proposed GA is capable of producing near-optimal solutions, making it an effective approach for cross-sector commodity portfolio optimization.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Komoditas Global Lintas Sektor, Genetic Algorithm, Mean-Variance Markowitz, Optimasi Portofolio, Kendala Cross-Sector Global Commodities, Genetic Algorithm, Mean–Variance Markowitz, Portfolio Optimization, Constraints
Subjects:	Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA402.5 Genetic algorithms. Interior-point methods.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Novita Indah Sari
Date Deposited:	31 Jul 2026 03:45
Last Modified:	31 Jul 2026 03:45
URI:	http://repository.its.ac.id/id/eprint/140938

Actions (login required)

View Item