Higher-order moment fuzzy portfolio selection model based on neural networks: Integrating behavioral finance and complex system analysis

In big data-driven socio-economic systems, traditional financial theories struggle to capture the complexity of investor behavior, uncertainty, and system dynamics. This paper introduces a high-order moment fuzzy portfolio selection model that integrates regret theory with a neural network-based evaluation approach. The model adopts a behavioral finance perspective by incorporating emotional factors such as regret and elation into the decision-making process. Trapezoidal fuzzy numbers are used to represent asset returns in order to account for uncertainty, while mean, variance, and skewness are included as optimization objectives. A neural network mutual evaluation framework assesses decision-making unit efficiency, using current ratio, quick ratio, cash ratio, and ownership ratio as inputs, and earnings per share, total revenue, net profit, and return on equity as outputs. The model supports mutual evaluation in fuzzy environments and helps select portfolios aligned with investors’ psychological expectations. Empirical analysis based on Chinese stock market data confirms its feasibility, robustness, and ability to reflect behavioral and systemic features in complex financial settings. This research bridges behavioral finance, fuzzy optimization, and complex systems theory, contributing to interdisciplinary development.