【作者】 单单；

【导师】 曹国华；

【作者基本信息】 重庆大学 ， 技术经济及管理， 2014， 博士

【摘要】 本文的主要研究内容是证券投资组合理论，针对证券市场会重复出现某些技术形态以及投资者会采用止损策略这两个现象，从如何确定股票收益率，如何建立合适的投资组合模型，如何设计高效的求解算法，止损策略如何对投资组合模型产生影响，如何确定止损策略的最优止损点和止盈点这五个方面入手，运用理论分析和定量分析相结合的方法做出了深入研究。主要研究内容可概述如下：①大量的理论和实践表明，股票市场会重复出现某些技术形态，具有某些形态的股票容易上涨，具有某些形态的股票容易下跌。如果投资者能够总结这些重复出现技术形态的规律并将其用于选择股票，则可以提升其在股市的盈利表现。针对这一现象，本文首先给出了一种新的量化股票收益率的方法并用双随机变量来描述股票收益率；随后，在已有文献的基础上建立了双随机安全第一投资组合模型并设计了一种融合了双重随机模拟技术与遗传算法的混合智能算法；最后，借助两类数值例子验证了新模型和算法的有效性，并根据数值结果给投资者提出相应的投资建议。②为了更好地求解双随机安全第一投资组合模型，本论文还设计了一种新的融合了LGMS-FOA算法和双重随机模拟技术的混合智能算法。首先，本文检验了果蝇算法（FOA）求解复杂优化问题的能力；随后，在FOA算法的基础上提出了LGMS-FOA算法；最后，将LGMS-FOA算法与双重随机模拟技术相结合得到了新算法，并将其与已有的算法进行比较。③人们的投资过程包括两部分，一是如何买入，二是如何卖出。然而研究卖出策略对投资组合模型影响的文献很少。止损策略是一种非常有效的卖出策略，本论文首先研究了止损策略对双随机安全第一投资组合模型的影响；随后，建立了一个带有止损策略的双随机安全第一投资组合模型并设计了一种融合了LGMS-FOA算法与双重随机模拟技术的混合智能算法；最后，给出了一个数值例子以验证模型和算法的有效性。基于以上研究，本文的结论如下：①本论文用双随机变量来量化股票收益率，该变量可以很好地体现技术形态与投资者异质性的特点；本论文建立的双随机安全第一投资组合模型既可以兼顾风险和收益，又可以适用于所有投资者，且数值例子也证明了本文模型和算法的有效性。②本文通过实验发现：已有的FOA算法不能很好地求解复杂优化问题，原因是FOA算法存在一种非线性的候选解产生机制，正是这种机制限制了FOA算法求解复杂优化问题的性能。为了克服FOA算法的缺陷，本文提出了LGMS-FOA算法，并从理论和实例两个方面证明了LGMS-FOA算法要优于FOA算法，同时，在求解双随机安全第一投资组合模型时，融合了LGMS-FOA和双重随机模拟技术的智能算法二也优于融合了遗传算法和双重随机模拟技术的智能算法一。③本文发现：止损策略会改变投资组合的比例，当设置止损点和止盈点后，投资者需要根据止损点和止盈点来改变相应的股票收益率，否则会造成投资组合模型失效。④在给定股票收益率以及风险和收益的条件下，选取合适的止损点和止盈点，止损策略的表现要优于非止损策略。这是因为采取止损策略的投资者关心的是股价超过止损点和止盈点的累计概率，而不再是更高的收益率，因而会采用新的资产组合比例。更多还原

【Abstract】 This paper studies the portfolio theory. There are two phenomena, the first is thatsome technical patterns occur repeatedly in stock market, the second is that investorsoften use exit strategy. In order to do research into the two phenomena, this paperstudy how to estimate the stock return, how to choose a portfolio model, how to designa hybrid intelligent algorithm, how to study the effect of exit strategy on portfoliomodel, how to determine the optimal stop-loss point and the stop-profit point. Themain contributions and originality contained in this dissertation are as follows:①Some technical patterns occur repeatedly in stock market. Stock with sometechnical patterns are much more likely to profit than those with other technicalpatterns. If investors can distinguish the recurring technical patterns, summarize thelaws, and use them to choose stocks, then it can greatly improve the performance ofprofit. Firstly, this paper propose a new method to estimate the stock return and usebirandom distribution to denote the final stock return, secondly, this paper build abirandom safety-first model based on the existing literatures and design a hybridintelligent algorithm integrating genetic algorithm and birandom simulation, finally, inorder o verify the validity of the model and the algorithm, this paper use threenumerical example to simulate the process of different investors and put forward thecorresponding suggestions to investors according to the numerical results.②In order to better solve the birandom safety-first model, this paper design anew hybrid intelligent algorithm integrating LMGS-FOA and birandom simulation.Firstly, this paper study the performance of FOA to address the complex optimizationproblems, secondly, LGMS-FOA is proposed based on FOA, finally, a new hybridalgorithm integrating LGMS-FOA and birandom simulation is proposed and comparedwith the existing algorithms.③As we all know, there are two important parts in investment process: how tobuy and how to sell. However, to the best of our knowledge, there is no research on theeffect of selling strategy on portfolio selection. Exit strategy is an effective sellingstrategy. Firstly, this paper innovatively study the effect of exit strategy on thebirandom safety-first model, secondly, this paper build a birandom safety-first modelbased on the exit strategy and design a hybrid intelligent algorithm integratingLGMS-FOA and birandom simulation, finally, a numerical example is proposed. Based on the above research, the conclusion of this paper is as follows:①Using birandom distribution to describe the stock return not only demonstratesthe features of technical patterns but also reflectes the investors’ heterogeneity. Thebirandom safety-first model not only considers the effect of profit and disaster but alsoapply to all investors through parameters adjustment. The numerical examples verifythe validity of the model and algorithm.②Through the analysis, it is found that FOA can not solve the complexoptimization problems effectively. This is because that FOA has a nonlinear generationmechanism of candidate solution which limit the performance of FOA. In order toovercome the disadvantages of FOA, LGMS-FOA is proposed and we theoreticallyprove that LGMS-FOA is better than FOA and the experimental results support ourconclusion. Meanwhile, when solving the birandom safety-first model, the hybridalgorithm integrating LGMS-FOA and birandom simulation is better than the hybridalgorithm integrating GA and birandom simulation.③Through the numerical example, this paper find that the exit strategy affectsthe buying strategy and the investor should adjust the stock return according to thestop-loss point and the stop-profit point, otherwise the portfolio model will becomeinvalid.④When the stock return, the risk and profit are given, the performance oftaking the exit strategy is better than when the exit strategy is not taken, if the stop-losspoint and the stop-profit point are appropriately set. The reason is that investors usingexit strategy will adopt a new portfolio proportion because he is concerned with theaccumulated probability of stock price exceeding the stop-profit point and the stop-losspoint not the higher stock return any more.更多还原