Research On Multi-period Portfolio Selection Model And Algorithm Based On Creditability Theory Considering Background Risk

The most important problem in portfolio theory is how to make the optimal proportion of financial products to achieve the maximum return and minimum investment risk.At present,most research focuses on the risk of investment,which is the uncertainty of the return of investors’ investment in the financial market.However,in real life,investors not only face financial risk,but also face background risk,such as the loss of human capital,pension portfolio risk,unexpected expenses related to health,labor income risk and real estate investment risk.Background risk comes from non-financial markets.It cannot be dispersed through portfolios in financial markets.The existence of background risk will directly affect investors’ investment activities in the financial markets,change the proportion of investors’ allocation to various risk assets,and affect investors’ investment decisions.If we only consider the risk brought by investors holding financial assets,and ignore the background risk in the non-financial market,the research of portfolio selection will be lack of scientific science and practicability due to the lack of comprehensive consideration of the actual conditions.The ignorance of background risk can cause portfolio failures,even lead to investors’ asset losses.Investors tend to conduct multi-period investment activities.In the investment environment,except the random uncertainty,there are also many fuzzy uncertainties.With the extensive attention to fuzzy set theory,some researchers began to analyze the fuzzy uncertainty on the basis of fuzzy set theory.However,most research are in single-period portfolio selection.Moreover,there are few research in multi-period credibility portfolio selection.Therefore,based on the credibility theory and considering background risk,this thesis studies how to build several multi-period portfolio selection models,and make in-depth analyses of the portfolio considering background risk and its pareto front.The main research work of this paper are as follows.(1)This thesis builds two credibility multi-period portfolio selection models with background risk appetite,the mean-lower absolute deviation-Burg’s entropy model and the mean-lower absolute deviation-Burg’s cross entropy model,discusses the changes of portfolio under different background risk appetite.Investors reduce the risk of investment through efficient diversification.They are more concerned about the downside risk.In the framework of credibility theory,lower absolute deviation is used to measure the risk,and the Berg’s entropy and the Berg’s cross entropy are used to measure the degree of diversification,respectively.We build the mean-lower absolute deviation-Burg’s entropy model and the mean-lower absolute deviation-Burg’s cross entropy model.The two multi-objective models above are transformed into two single-objective models,and the dragonfly algorithm is used to solve the proposed models.Empirical analyses show that background risk appetite has an impact on the portfolio selections.Investors asset ratio is different when investors focus on different goals.(2)This thesis proposes a credibility mean-semi entropy multi-period portfolio selection model considering background risk and designs a multi-objective hybrid dragonfly algorithm to solve the proposed model.Under the premise of considering liquidity constraints,we maximize returns and minimize downside risks,propose a credibility mean-semi entropy multi-period portfolio selection model considering background risk.In the proposed model,some realistic factors are considered,such as cardinality constraints,risk-free assets,non-short selling,transaction costs,etc.The traditional dragonfly algorithm is improved by optimizing the parameters,constrained non-dominant sorting and external storage archives,and combined with non-dominant sorting genetic algorithm II,a multi-objective hybrid dragonfly algorithm is proposed.The experiments show that proposed algorithm has good performance in diversity distribution and convergence,and the model is effective under different cardinality constraints.The pareto front of the model moves to the upper right,and the overall risk of investors increases accordingly,while the background risk increases.(3)This thesis provides a credibility mean-Shannon entropy multi-period portfolio selection model with investor sentiment and background risk.Previous financial theories assum that investors were immune to emotions and able to make completely rational investment choices.However,many empirical studies have found that investors tend to have positive and negative emotions.Considering the background risk,investor sentiment is introduced to establish the credibility mean-Shannon entropy multi-period portfolio selection model.In order to solve the proposed model,we improve the proposed algorithm by harmony average distance.The results of the experiment show that the difference of investor sentiment will result in the different portfolio results.When investors maintain a positive attitude,they will overestimate the return on investment and ignore the investment risk.When investors remain negative,they overestimate the risks and underestimate the returns.