本文選題：指數(shù)跟蹤　切入點：FastICA　出處：《華東師范大學》2014年碩士論文

【摘要】：隨著證券市場的發(fā)展和演變,市場指數(shù)體系的建立漸漸趨于完善。指數(shù)化投資理念已成為世界范圍內的主要投資策略和投資方法之一,作為指數(shù)化投資的重要技術,指數(shù)跟蹤組合復制方法得到了國內外眾多學者的關注和研究。但是大部分研究都是以證券收益率為出發(fā)點來構造指數(shù)跟蹤模型,很少會對股票價格的時間序列做研究。在指數(shù)跟蹤投資組合的成分股的選擇方法上,大多數(shù)學者采用的是市值排名法、按行業(yè)抽樣或者簡單的隨機抽樣,這些選股方法都缺乏對股票以及目標指數(shù)的歷史信息進行充分的挖掘和利用。另外,已有研究基本上都只對一種復制方法進行研究,缺乏對不同跟蹤組合復制方法的跟蹤效果作系統(tǒng)的比較分析。本文針對以上問題提出了一個新的跟蹤組合復制的方法,以價格的時間序列為研究對象,用FastICA方法對股票價格時間序列進行特征提取,根據(jù)不同的趨勢特征對股票進行聚類分層,充分挖掘股票和目標指數(shù)的歷史信息。然后提出了基于聚類的Non-negative LARS算法進行選股；最后,對選入的股票按照目標函數(shù)優(yōu)化求解,以此賦予股票不同的權重,從而得到復制跟蹤的最優(yōu)組合。在實證分析部分,本文還對各種指數(shù)跟蹤組合復制方法進行了多角度的對比,結果顯示本文提出的指數(shù)跟蹤復制方法在樣本內的跟蹤效果是相對最優(yōu)的,而且基于ICA的聚類和基于聚類的Non-negative LARS算法也取得了不錯的效果；在樣本外的短期和中期,該方法也有很好的跟蹤效果。另外,在實證分析時還嘗試從更大的樣本空間挑選股票構造跟蹤組合,結果表明從滬深300指數(shù)的成分股挑選股票構造的跟蹤組合其效果要優(yōu)于從所有股票中選股構造的跟蹤組合,這對指數(shù)跟蹤技術的一項探索,也是一項有益的嘗試,有一定的實際參考價值。
[Abstract]:With the development and evolution of securities market, the establishment of market index system tends to perfect gradually.The concept of indexed investment has become one of the main investment strategies and investment methods in the world. As an important technology of indexed investment, the method of index tracking portfolio replication has been paid attention to and studied by many scholars at home and abroad.In the selection method of index tracking portfolio composition, most scholars use market value ranking method, sampling by industry or simple random sampling.These stock selection methods are short of the historical information of stock and target index.In addition, most of the previous studies focus on only one replication method, which lacks a systematic comparative analysis of the tracking effects of different tracking combinations.In order to solve the above problems, a new method of tracking portfolio replication is proposed in this paper. Taking the time series of price as the research object, the FastICA method is used to extract the features of the time series of stock prices.Then we propose a clustering based Non-negative LARS algorithm for stock selection. Finally, we optimize the stock selection according to the objective function to give different weights to the stock and obtain the optimal combination of replication tracking.In the part of empirical analysis, this paper also makes a multi-angle comparison of various index tracking combination replication methods. The results show that the index tracking replication method proposed in this paper is relatively optimal in the sample tracking effect.Moreover, the clustering algorithm based on ICA and Non-negative LARS algorithm based on clustering have also achieved good results, and the method also has good tracking effect in the short and medium term outside the sample.In addition, in the empirical analysis, we also try to select the stock structure tracking portfolio from a larger sample space.The results show that the tracking combination selected from the constituent stocks of CSI 300 index is more effective than the one from all stocks, which is also a useful attempt to explore the technology of index tracking.It has certain practical reference value.
【學位授予單位】：華東師范大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：F832.51;F224

【參考文獻】

相關期刊論文 前5條

1 李儉富;馬永開;曾勇;;指數(shù)跟蹤研究評述[J];當代經(jīng)濟管理;2005年06期

2 嚴武,洪道麟;我國指數(shù)基金運作與績效研究[J];統(tǒng)計研究;2000年12期

3 倪蘇云,吳沖鋒;跟蹤誤差最小化的線性規(guī)劃模型[J];系統(tǒng)工程理論方法應用;2001年03期

4 李儉富;馬永開;;基于證券價格時間序列的協(xié)整優(yōu)化指數(shù)跟蹤方法研究[J];系統(tǒng)工程理論與實踐;2006年11期

5 張玲;ETFs跟蹤誤差產(chǎn)生原因探討[J];證券市場導報;2002年11期

，

本文編號：1699815