Analysis Of Investments In Securities Based On Neural Networks And Genetic Algorithms

Securities business is one investment market with high risk and high income. To earn large but with low risk is the dream of all the investors. To realize this dream, two questions must be concerned. The first is which security should be chose and the second is how to complete portfolio. Based on tradition and innovate theory, this dissertation attempts to deal with these two problems with advanced technical means.Plenty researchers worldwide try to forecast security market, but if security market is efficient all the forecast are null. From both theory and practice the dissertation draw the conclusion that our country's security market is inefficient, security market can be forecasted at least during short terms. Being affected with too much factors, security forecast is very difficult. The dissertation explores to use nonlinearity means– BP neural network to solve this problem. By practice, the established BP model gets good result and is efficient. What is more, the model can be changed according to investors'favor and exercises.On the basis of Markowitz portfolio investment model, widely used risk tool VaR in investment field is applied to establish a model with one objective function of profit covariance and multi-constraint of VaR and profit rate. Under the hypothesis of normal distribution,the model non-linear constraint is deeply simplified. Then aiming at this model, genetic algorithm is designed for solving it. The optimum solution in the simulation about this portfolio model is given with algorithm. Its result is good not only in fitting in the VaR limitation but also catering for different investors with different profit needs.This dissertation combines security selection and portfolio together to form a relatively complete investment system. This system eases the process of investment especially for the layman of security investment. Practice proves this system is feasibility and maneuverability and can...