Research On Theoretical Model And Algorithm Of Portfolio Selection

The existed Theory of Portfolio Selection is analyzed; the objective function and subject conditions are developed to fit the investment of real market; the model of Portfolio Selection is solved by heuristic algorithms.The interval constraint is added to the mean-variance model. This non-linear programming model is solved by genetic algorithm. The initial population is generated randomly. The best unit protection is applied to selection. The arithmetic crossover is used in crossover operator. The constraints are tested after mutation.The risk probability based on Value at Risk is used as the objective function. The stochastic programming model is constructed. The data of investment proportion and risk probability is generated from stochastic simulation. The risk probability function is approximated by BP neural network. The stochastic programming model is solved by genetic algorithm.The transaction cost is added to the model of portfolio selection. The objective function is the combination of expected revenue and variance. The risk free assets are also added to the model. The coefficient of the risk is introduced to reflect inclination of the investor. The probability of crossover and mutation is adaptive in the genetic algorithm.