Study On The Selection Of Risky Assets Based On Attribute Reduction Theory

Capital market is a market where benefits and risks coexist.The key issue that the investors are confronted with is how to avoid risks while obtaining benefits.The investors,especially the institutional investors,usually select a variety of risky assets to construct the portfolio to dissolve risks.While,the core is to screen out the potential risky assets among thousands of traded assets to construct the portfolio.This paper takes the stocks for example,uses the attribute reduction theory to choose the minimal attribute sets that identifies the risk assets,which are used to categorizes the risky assets by cluster analysis technique.Clusters with better performance are selected to construct stock asset pool and the investment performance of the selected stocks will be further tested against the stock index.Chapter 1 introduces the significance of this research and surveys important literatures related to the portfolio models and algorithms of attribute reduction,and also outlines the contents,designs,and innovations of this research.In Chapter 2,the attribute reduction theories and algorithms are investigated,and then an improved algorithm of attribute reduction is proposed based on attribute multiplicity.This algorithm takes the attributes multiplicity between samples as the basis of discretization and unchanged relation of unidentifiability of information system as the basis of attribute reduction,which remedies the deficiencies of previous algorithms.Chapter 3 selects these stocks included in HS300 that survived before 2005 to till now to address the implementation of attribute reduction algorithm.First,those stocks with better performance are selected by industries,and a variety of indices reflecting stock returns and company's financial statues are picked out;then the minimal attribute sets of different industries are computed based on the improved attribute reduction algorithm,which is used to cluster the selected stocks by industries.And the clusters with better performance will be selected to construct the assets pool.Chapter 4 computes the optimal portfolios based on Monte Carlo stochastic simulation and rolling Mean-CVaR model to test the performance of assets pool proposed by improved attribute reduction algorithm.Both methods validate the advantages of stocks selected by the improved attribute reduction algorithmChapter 5 summarizes the research work of this paper,and takes a prospect for future research.