Abstract: With VaR as a measure of investment combination‚the “Mean-VaR” optimal portfolio model was established by using Markowitz’s Portfolio Theory．Furthermore‚Mont Carlo simulation method as well as the Genetic Algorithm was designed for solving this model．Following‚through the study on China stock market analysis was given to examine the proposed method‚some valuable results that if the fatter the tail of the distribution is‚or the higher VaR confidence level is‚then the more conservative
the portfolio decision is can be obtained by discussing the effect on investment from the hypothesis of between the distribution of capital asset rate and the VaR confidence level．

Key words: value at risk, genetic algorithm, risk aversion, confidence level