The aim of the study is to prove that dynamic portfolios can effectively reflect the temporal dynamics of current risks of a higher order, providing greater reliability and stability compared to traditional portfolios. The subject is the economic imbalance in portfolio models, which occurs when different participants have different level of knowledge about market conditions and the performance of assets. In an environment where traditional portfolios have shown low returns due to the pronounced peaks and sharp declines in financial asset returns, as well as their inability to account for dynamic changes in financial risks. This study incorporates higher-order short-term risks into traditional portfolios in order to mitigate the effects of deviations from the normal distribution. The methodology is based on the concept of multiple financial time series and the VAR-ICA-MGARCH model. This model effectively captures the conditional mean, the covariance matrix, the mutual asymmetry matrix, and the mutual kurtosis matrix, thereby characterizing temporal changes at higher-order moments. Due to the inherent nonlinearities of dynamic portfolio optimization tasks, we use a genetic algorithm to solve the dynamic portfolio model. The results of the study show that dynamic portfolios can effectively reflect the changing dynamics of current higher-order risks, while providing greater reliability and stability than traditional portfolios. Even when exposed to such complex risks, dynamic portfolios perform better. The practical significance of this research lies in determining the time-varying weighting coefficients for the portfolio and conducting both simulation experiments and empirical analysis.