Abstract: Two defects in traditional portfolio theory,without considering higher moments risk and settling problem staticly,are pointed out in the paper.To measure higher moments risk,the multivariate GARCHSK model is established. Then,based on Taylor series expansion of utility function,the dynamic portfolio model with higher moments risk is derived,and the model is solved by Genetic Algorithm. Empirical results show that not only higher moments risk exists in Chinese stock markets,but also the risk has time varying character. It is necessary for the investors to change their portfolio weights to avoide higher moments risk.

Key words: higher moments risk, dynamic portfolio, multivariate GARCHSK model, Genetic Algorithm