本文關(guān)鍵詞： 投資組合 Black-litterman模型 Bootstrap方法 BP神經(jīng)網(wǎng)絡(luò)　出處：《大連理工大學(xué)》2013年碩士論文　論文類型：學(xué)位論文

【摘要】：我國證券投資基金行業(yè)十五年來的發(fā)展令人矚目,然而最近幾年,受到次貸危機(jī)和歐債危機(jī)的影響,證券市場產(chǎn)生了大幅波動。在風(fēng)險增加的情況下,對證券投資組合進(jìn)行研究成為對基金管理的必然要求。有關(guān)研究表明,資產(chǎn)的有效配置對投資業(yè)績的貢獻(xiàn)率高達(dá)93.6%。近年來,國內(nèi)基金行業(yè)開始對數(shù)量化資產(chǎn)配置模型展開研究,這些研究大多是吸收了國外的數(shù)量化模型,但中國市場畢竟有別于國外市場,因此,有必要探索改進(jìn)國外成熟的數(shù)量化模型,并應(yīng)用于中國證券市場的方法。 Black-litterman模型是由高盛公司提出的,該模型從誕生之初就被實際應(yīng)用于基金投資決策中,經(jīng)過多年的發(fā)展已經(jīng)得到廣泛的認(rèn)可。本文首先對經(jīng)典的均值方差理論作了簡單的回顧,然后對原始Black-litterman模型中各個復(fù)雜的輸入?yún)?shù)作了詳細(xì)的論述,并且使用了Bootstrap方法和神經(jīng)網(wǎng)絡(luò)對原模型做了改進(jìn),建立了新的ANNs-BLR模型。Bootstrap方法能夠很好地修正原模型在計算市場均衡收益率時產(chǎn)生的誤差；神經(jīng)網(wǎng)絡(luò)模型能夠很好地捕捉到股票市場復(fù)雜的波動規(guī)律,以該模型的預(yù)測結(jié)果替代原模型中投資者的主觀觀點(diǎn),可以有效的提升模型績效。 本文選取了剔除數(shù)據(jù)不全股票后的上證50指數(shù)的41只權(quán)重股作為樣本。在采用了誤差修正模型,并使用BP神經(jīng)網(wǎng)絡(luò)的估計量替代模型觀點(diǎn)收益向量后,得到了下一期的模型收益和資產(chǎn)配置結(jié)果。在第一部分實證研究中,結(jié)果表明,ANNs-BLR模型無論是在有無賣空限制、有無配置權(quán)重上限的情況下,配置的穩(wěn)定性都優(yōu)于均值方差模型,模型收益率和夏普比率也都高于均值方差模型。在第二部分實證中,本文針對中國股票市場非流通股票普遍存在的情況,對分別采用總市值權(quán)重、流通市值權(quán)重以及經(jīng)過誤差調(diào)整的模型配置績效進(jìn)行了比較,結(jié)果顯示流通市值權(quán)重更加適合該模型,能夠取得更好的投資績效；而且誤差調(diào)整過程也能夠有效改善模型績效。
[Abstract]:The development of China's securities investment fund industry in the past 15 years has attracted great attention. However, in recent years, affected by the subprime mortgage crisis and the European debt crisis, the securities market has produced large fluctuations. Research on the portfolio of securities has become an inevitable requirement for fund management. Related studies show that the contribution rate of effective asset allocation to investment performance is as high as 93.6. In recent years, the domestic fund industry has begun to study the quantitative asset allocation model. Most of these studies have absorbed the quantitative models of foreign countries, but the Chinese market is different from foreign markets after all. Therefore, it is necessary to explore and improve the mature quantitative models abroad and apply them to the Chinese securities market. The Black-litterman model was put forward by Goldman Sachs, which was applied to fund investment decision from the beginning of its birth. After years of development, it has been widely accepted. In this paper, the classical mean-variance theory is reviewed briefly. Then the complicated input parameters in the original Black-litterman model are discussed in detail, and the Bootstrap method and neural network are used to improve the original model. A new ANNs-BLR model. Bootstrap method is established to correct the error of the original model in calculating the market equilibrium return rate, and the neural network model can capture the complex volatility law of the stock market. The prediction results of the model can effectively improve the performance of the model by replacing the subjective point of view of the investors in the original model. In this paper, 41 weight shares of Shanghai 50 index after eliminating incomplete data are selected as samples. After adopting the error correction model and using the BP neural network estimate to replace the model income vector, In the first part of the empirical study, the results show that the stability of the model is superior to that of the mean variance model, regardless of whether there is a short selling limit or not, with or without the upper limit of the allocation weight, and the results of the first part of the empirical study show that the stability of the model is better than that of the mean variance model. The return rate and Sharp ratio are also higher than the mean variance model. In the second part of the empirical analysis, this paper uses the total market value weight to deal with the prevailing situation of non-circulating stocks in Chinese stock market. The results show that the weight of circulation market value is more suitable for the model and can achieve better investment performance, and the error adjustment process can also effectively improve the model performance.
【學(xué)位授予單位】：大連理工大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2013
【分類號】：F832.51;F224

【參考文獻(xiàn)】

中國期刊全文數(shù)據(jù)庫 前10條

1 劉華;;基于有效市場理論的我國股市效率研究[J];財會通訊;2012年23期

2 郭暢;;基于Black-Litterman模型的資產(chǎn)配置研究[J];湖北大學(xué)學(xué)報(自然科學(xué)版);2009年01期

3 陳夢根;中國股市長期記憶效應(yīng)的實證研究[J];經(jīng)濟(jì)研究;2003年03期

4 王晟;蔡明超;;中國居民風(fēng)險厭惡系數(shù)測定及影響因素分析——基于中國居民投資行為數(shù)據(jù)的實證研究[J];金融研究;2011年08期

5 王敬;王穎;;機(jī)構(gòu)投資者資產(chǎn)配置方法研究[J];價值工程;2006年02期

6 溫琪;陳敏;梁斌;;基于Black-Litterman框架的資產(chǎn)配置策略研究[J];數(shù)理統(tǒng)計與管理;2011年04期

7 孟勇;;對Black-Litterman模型加入主觀收益方法的改進(jìn)[J];統(tǒng)計研究;2012年02期

8 盧方元;中國股市收益率的多重分形分析[J];系統(tǒng)工程理論與實踐;2004年06期

9 何建敏,常松;中國股票市場多重分形游走及其預(yù)測[J];中國管理科學(xué);2002年03期

10 蘇治;方明;李志剛;;STAR與ANN模型:證券價格非線性動態(tài)特征及可預(yù)測性研究[J];中國管理科學(xué);2008年05期

中國博士學(xué)位論文全文數(shù)據(jù)庫 前4條

1 蔡志岳;中國上市公司信息披露違規(guī)的動因、市場反應(yīng)與預(yù)警研究[D];廈門大學(xué);2007年

2 佟偉民;股指期貨交易中操縱行為識別方法研究[D];哈爾濱工業(yè)大學(xué);2008年

3 孫曉琳;基于Kalman濾波和BP神經(jīng)網(wǎng)絡(luò)的財務(wù)危機(jī)動態(tài)預(yù)警模型研究[D];哈爾濱工業(yè)大學(xué);2010年

4 劉德紅;基于微觀結(jié)構(gòu)理論的證券市場可預(yù)測性研究[D];北京交通大學(xué);2012年

，

本文編號：1515676