Topology-Informed Financial Network Approach to Portfolio Optimization Using Fuzzy Decision-Making and Genetic Algorithms: Evidence from the Istanbul Stock Exchange

This study proposes a hybrid portfolio optimization framework integrating financial network analysis, Adaptive Neuro-Fuzzy Inference Systems (ANFIS), and GA for asset allocation in BIST. The empirical analysis focuses on constituent firms within the BIST 30, BIST 50, and BIST 100 indices using daily stock market data covering the period 2000–2025. Financial network centrality indicators and technical analysis variables were employed to identify structurally influential assets and model nonlinear investment decision dynamics under market uncertainty. The ANFIS framework was utilized to capture complex relationships between technical indicators and portfolio allocation decisions, while Genetic Algorithms optimized portfolio weights under return maximization and downside-risk minimization constraints. To reduce overfitting risk, Principal Component Analysis (PCA) and K-fold cross-validation procedures were implemented during model training. The proposed framework was additionally evaluated using out-of-sample backtesting over the 2021–2024 period and compared against benchmark portfolio strategies, including Equal Weight and Minimum Variance portfolios. Empirical findings indicate that the ANFISGA framework achieved superior risk-adjusted performance, higher Sharpe and Sortino ratios, and lower maximum drawdown during volatile market conditions. The study contributes to the portfolio optimization literature by integrating financial network indicators with adaptive fuzzy decision systems and evolutionary optimization techniques within an emerging market context. The proposed framework is intended primarily as an adaptive portfolio decision-support system rather than a purely predictive forecasting model.