The CVaR Portfolio Investment Model Based On Genetic Algorithm

The methods and theories used to research on risk measurement and optimization of securities portfolio have significant meaning as to finance risk control and finance order stabilization. Portfolios theory, one of the important research contents in Economics, aims to attain the portfolios of the maximum of investment’s return with the given value of the risk of portfolios, or of the minimum of investment’s risk with the given level of the investment’s return. New risk measurement methods VaR (value-at-risk) and CVaR (conditional value-at-risk) are put forth recently. Because of its eminent properties, CVaR becomes a latest research content in risk management.The research in this paper is based on the framework of Markowitz portfolio optimization theory. Under this framework, this paper briefly introduces theoretical foundation of modern portfolio and the mean-variance model. Then, on the foundation of the definition of one of the most important methods of risk measurement VaR, this paper puts forward CVaR, a risk measurement tool which has superior properties, and introduces a model which connects VaR and CVaR in one equation. In addition, considering the influence placed by transaction cost on the returns of portfolio, the Mean-CVaR model with the restriction of transaction cost is studied in this paper. The Genetic Algorithms were designed, which were used to solve the model and gave the number imitations of the model. At last, by using Matlab and Excel, this paper makes empirical research on portfolio and efficient frontier of Mean-CVaR Model, and then systematically researches the influence of constraint conditions--confidence level and transaction cost.