本文關(guān)鍵詞：基于多準(zhǔn)則決策組合優(yōu)化拓展Black-Litterman模型的應(yīng)用　出處：《復(fù)旦大學(xué)》2012年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 資產(chǎn)配置 Black-Litterman模型 多準(zhǔn)則決策 遺傳算法

【摘要】：Markowitz在1952年提出基于均值-方差模型的投資組合理論,使資產(chǎn)配置跨入了數(shù)量化的時(shí)代,但是這個(gè)模型不能很好地融合投資者的主觀觀點(diǎn)。而1992年開始為人所知的Black-Litterman資產(chǎn)配置模型在半強(qiáng)式有效市場(chǎng)的假設(shè)基礎(chǔ)上,結(jié)合使用了資本資產(chǎn)定價(jià)模型和貝葉斯規(guī)則,可以很方便地融入了投資者的主觀觀點(diǎn)。本文應(yīng)用針對(duì)市場(chǎng)變量的Black-Litterman模型(BLm模型)來獲得A股行業(yè)指數(shù)超額收益率的后驗(yàn)分布。 現(xiàn)有的關(guān)于Black-Litterman模型的研究,常常在組合優(yōu)化階段采用Markowitz的臨界線算法。這個(gè)優(yōu)化方法假設(shè)所有投資者在市場(chǎng)均衡時(shí)使用同樣的偏好,不能融合投資者個(gè)人偏好。為了處理個(gè)人偏好,本文使用了多準(zhǔn)則決策(MCDM)方法進(jìn)行組合優(yōu)化。 本文利用加權(quán)法把多目標(biāo)優(yōu)化問題轉(zhuǎn)化為單目標(biāo)優(yōu)化問題。首先使用插值法把投資者的對(duì)各個(gè)目標(biāo)的偏好轉(zhuǎn)化為具體的效用函數(shù),然后把各個(gè)效用函數(shù)整合成一個(gè)總體效用函數(shù),作為優(yōu)化問題的目標(biāo)函數(shù)。在優(yōu)化中采用遺傳算法尋優(yōu),較好的解決了組合優(yōu)化問題中的NP完全問題和函數(shù)搜索空間復(fù)雜(非凸、非凹)的問題。 此外,本文驗(yàn)證了在MCDM的組合優(yōu)化模型中,Black-Litterman模型的魯棒性極其有限。提升資產(chǎn)配置的魯棒性需要依靠添加約束條件,并采用適當(dāng)?shù)膯l(fā)式算法來進(jìn)行優(yōu)化。
[Abstract]:In 1952, Markowitz put forward the portfolio theory based on mean-variance model, which made asset allocation enter the era of quantification. However, this model does not integrate investors' subjective views very well. And the Black-Litterman asset allocation model, known since 1992, is based on the assumption of semi-strong efficient markets. . A combination of capital asset pricing model and Bayesian rules is used. This paper applies the Black-Litterman model to market variables and blm model. To obtain the A-share industry index excess return posterior distribution. The existing research on Black-Litterman model. Markowitz's critical line algorithm is often used in portfolio optimization, which assumes that all investors use the same preference in market equilibrium. In order to deal with individual preferences, this paper uses the multi-criteria decision making (MCDM) method to optimize the portfolio. In this paper, the weighted method is used to transform the multi-objective optimization problem into a single-objective optimization problem. Firstly, the interpolation method is used to convert the investor's preference to each objective into a specific utility function. Then the utility function is integrated into a total utility function, which is the objective function of the optimization problem. Genetic algorithm is used to optimize the optimization. NP-complete problems in combinatorial optimization problems and complex (non-convex, non-concave) function search spaces are well solved. In addition, this paper verifies that the robustness of Black-Litterman model is very limited in MCDM's combinatorial optimization model. To improve the robustness of asset allocation, we need to add constraints. An appropriate heuristic algorithm is used for optimization.
【學(xué)位授予單位】：復(fù)旦大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2012
【分類號(hào)】：F830.91;F224

【引證文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前1條

1 袁順;基于Black-Litterman模型的海洋災(zāi)害補(bǔ)償基金收益研究[D];中國海洋大學(xué);2014年

，

本文編號(hào)：1427502