Research Of Portfolio Investment Based On Improved Particle Swarm Optimization

With the development of China's economy,the increase of household income and the improvement of financial system,investment has become the main solution to maintain and add the value of assets.The stock market is high-risky,and many uncertain factors make it difficult to predict the market trend accurately.How to ensure that the investment funds allocated to a variety of assets under the premise of obtaining the expected income to determine the optimal weight has become the focus attention of investors,thus developing the portfolio theory.Since Markowitz put forward the MV model,scholars have been studying the risk measurement which can replace the variance better and the portfolio model which is more suitable for market conditions.Based on the mean-CVaR model,the adjusted mean-CVaR extended model and the mean-CVaR-Tsallis generalized entropy expansion model are established in this thesis.The improved particle swarm optimization algorithm is designed to find the algorithm with stronger searching ability and cango out from the local convergence.These topics not only has certain theoretical significance but also with practicability.This thesis first studies classic portfolio theory such as particle swarm optimization theory and entropy theory.Then,based on the actual market environment,a mean-CVaR extended model with transaction costs,industry constraints and investment ratio constraints is established.On the basis of this,the thesis introduces the modified coefficient to adjust the loss function in the model,thus establishing the mean-CVaR extended model in which the loss function is a general nonlinear condition.It is difficult to solve nonlinear model directly.This thesis designs improved particle swarm optimization algorithm which introduces predation strategy,hybrid strategy,adjusted inertia weight to avoid the algorithm falling into local optimum prematurely and improve searching ability.Then the empirical analysis of the SSE 50 Index stock is carried out.With the cross-integration of knowledge from different disciplines,some related concepts such as entropy have been introduced into the portfolio optimization model.The entropy value in physics is proportional to the uncertainty.If it is used in the portfolio model,it can show correlation degree between the assets in the portfolio,so entropy can measure the dispersion degree of the portfolio.Based on this background,a mean-CVaR-Tsallis generalized entropy extended model is established.The model uses CVaR and Tsallis generalized entropy as objective functions.To solve the model,an improved multi-objective particle swarm optimization algorithm is designed.Finally,the empirical analysis is carried out,which is compared with the adjusted mean-CVaR extended model and the the SSE 50 Index.The results of empirical analysis show that the mean-CVaR-Tsallis generalized entropy expansion model can not only reflect the principle of diversification and reduce the risk,but also has strong practicability.In most cases,the income performance is slightly better.