A Study Of Portfolio And Solving By Improved Bionic Algorithm

Owing to the intense impact of the world economy and financial sit-uation,more and more people choose to invest in stocks. Nowadys, portfo-lio investment has become a hot topic in real life and in literature. Bea-cause of the variety of stock types, the development potential, return and risk on investment are different. However, the investors' capital is fixed, so we need to consider the allocation of the fixed investment to various stocks. This dissertation addresses the optimal decision-making on port-folio investment by using bionic algorithm, in the domestic stock market environments.In this thesis, Markowitz portfolio theory and the real situation in the domestic stock market are fused. The entropy is combined with vari-ance to further measure the risk, and the expert evaluation is emplyed to make expected return vector and fuzzify the covariance matrix, so as to enhance the model feasibility. Considering the relevant policies, as well as other practical constraints, the maximum and minimum levels of in-vestment are set up, and an improved model is established. Furthermore, three improved biomimetic approaches are proposed to solve the new model.To overcome the slow convergence rate and long computation time as shortcomings of traditional genetic algorithms, a self-adaptive parallel genetic algorithm is designed. In the algorithm, a new mechanism of mul-ti-processor parallel processing is put forward, and crossover and muta-tion operators are modified with self-regulation. Finally, a numerical example is provided to demonstrate the effectiveness of the algorithm.The crossover-operation of the genetic algorithm is introduced to a standard particle swarm optimization (PSO) algorithm to make particles crossable selectively. A parallel self-adaptive vehicle is applied to enable the algorithm to run over multi-processors at the same time, and adjust the weights by itself The improved algorithm is compared with the stan-dard one through some empirical analysises,Noting that the genetic algorithm has the characteristics of strong global search but weak local optimization, while the back propagation (BP) neural network owns the opposite, in this thesis, a genetic algorithm and back propagation algorithm (GABP) is therefore given by combining the complementary advantages of both. The improved genetic algorithm is excuted for macro-search, and then the batch BP neural network with self-regulated learning rate for micro-adjustment to search the optimal solution. Finally, a numerical example is provided to test its feasibility.