Application Of CVaR Model Based On Genetic Algorithm To Portfolio

Financial risk management is an essential problem for financial practioners, academic scholars and related supervision departments. Value-at-Risk (VaR) is an important direction in recent financial research, it's a statistic method to measure the risk of stock markets, and it is used to predict the largest lost within a time period and portfolio. Due to the absence of some good properties such as sub-additivity and convexity, study on Conditional Value-at-Risk (CVaR) has attracted many researchers'interests in this field. CVaR is the expected losses or average losses exceeding VaR. It not only has the advantages of VaR model, but also is more practical and reasonable from the application point of view. Hence it has more and more applications. VaR and CVaR,especially the latter, have become standard tools of financial risk evaluation currently.The application of CVaR model to portfolio is researched in this thesis, and the main results obtained are as follows:1. Because the loss function in CVaR model is usually complex, it is rather difficult for a linear loss function to describe the loss accurately. So a nonlinear loss function is proposed, by which a nonlinear programming CVaR model is designed.2. An improved genetic algorithm is proposed and implemented in order to solve the new CVaR model. The result of the experiment shows that the new model can lower both VaR and CVaR, which means the risk is reduced apparently.3. The risk composition of portfolio is analyzed, and some risk analyzing methods, such as Margin CVaR, Constituent CVaR and Increment CVaR, are studied. Then a risk aversion model is designed. With this model, the CVaR is reduced and meanwhile the revenue of portfolio is maintained.