Gellysa Urva
Institut Teknologi dan Bisnis Riau Pesisir

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

Hybrid Genetic Algorithm and XGBoost Model for Optimizing Sharia Compliant Stock Portfolios in the Indonesian Capital Market Gellysa Urva; Welly Desriyati
Journal of Innovation Information Technology and Application (JINITA) Vol 8 No 1 (2026): JINITA, June 2026
Publisher : Politeknik Negeri Cilacap

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35970/jinita.v8i1.2999

Abstract

This study proposes a hybrid portfolio optimization model that integrates Genetic Algorithm (GA) and Extreme Gradient Boosting (XGBoost) to optimize Sharia-compliant stock portfolios in the Indonesian capital market. The model leverages the predictive capability of XGBoost and the adaptive optimization mechanism of GA to achieve an optimal balance between return and risk under Sharia investment principles.  The dataset consists of historical daily stock price data from six Sharia-compliant stocks (ADRO, INDF, KLBF, TLKM, UNVR, and ICBP) during 2023. XGBoost was employed to predict stock returns using selected technical indicators, and its performance was evaluated using Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The predicted returns were then used as inputs for portfolio optimization using GA.  The results show that the hybrid model produces an optimized portfolio with KLBF (22.7%) and ICBP (21.3%) receiving the highest allocations, indicating favorable risk–return characteristics. The model achieved RMSE values ranging from 0.0081 to 0.0142 and MAPE values between 4.24% and 6.11%, demonstrating high predictive accuracy. Furthermore, the hybrid portfolio outperformed benchmark strategies, achieving an expected return of 6.42% with a Sharpe Ratio of 0.30.  These findings confirm that the integration of GA and XGBoost enhances both prediction reliability and portfolio optimization performance. This study contributes to the development of artificial intelligence-based portfolio optimization in Islamic finance and provides a practical framework for ethical and data-driven investment decision-making.
Modification of the Proximity Indexed Value Method Based on Distance-Based Normalization to Improve the Stability of Alternative Rankings Desyanti Desyanti; Gellysa Urva; Sunggito Oyama; Kristina Febriani
Journal of Decision Support Systems and Multi-Criteria Decision Making Vol. 1 No. 3 (2027): July 2027
Publisher : Asosiasi Peneliti Informatika dan Komputer untuk Riset (PILAR)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.67449/jodesma.v1i3.16

Abstract

The Proximity Indexed Value (PIV) method provides a proximity-based mechanism for ranking alternatives in multi-criteria decision-making; however, its ranking results can be affected by the normalization process and the relative characteristics of criterion values. This study proposes a Distance-Based Proximity Indexed Value (D-PIV) method by incorporating distance-based normalization into the PIV framework to represent the proximity of alternatives to criterion-specific ideal conditions more explicitly. The proposed method was evaluated using an alternative location selection dataset comprising eight alternatives and six criteria, including one cost criterion and five benefit criteria. The evaluation included comparative ranking analysis using 36 systematic criterion-weight perturbation scenarios with changes of ±0.05, ±0.10, and ±0.15. The results show that D-PIV changes the ordering of Loc 3 and Loc 8 compared with conventional PIV, resulting in a ranking that fully corresponds to the reference ranking. The Spearman correlation increased from 0.9762 for PIV to 1.0000 for D-PIV. However, the sensitivity analysis indicates that the distance-based modification does not uniformly improve ranking stability across all alternatives. D-PIV increases the stability of Loc 8 from 69.44% to 72.22%, while lower stability is observed for Loc 3 and Loc 6 compared with PIV. These findings demonstrate that D-PIV contributes a reference-oriented normalization mechanism that improves baseline ranking consistency while redistributing, rather than uniformly eliminating, sensitivity to criterion-weight variations. The study therefore provides both a methodological contribution through the integration of distance-based normalization into PIV and an empirical contribution through systematic evidence of its effects on ranking agreement and stability.