Research On The Selection Of Risky Assets Based On Cluster Analysis Via Multi-indices Dynamic Time Warping Distance

The stock market is an important source to aggregate wealth while it is accompanied by huge risk.An investor has to care more about how to evade financial risks while achieving the expected return.It is a key issue for an investor to select the risky assets with investment potentials to construct a robust portfolio among more than 3000 listed companies in Shanghai and Shenzhen Stock Exchanges.In this paper,we first take the constituent stocks of HS300 as sample pool and choose different financial indices for each industry that reflect the market performances of companies in this industry.And the similarity between stocks for each index is computed by dynamic time warping distance to construct a warping distance matrix.Then K-Means cluster analysis algorithm is used to classify the stocks and a group of stocks with better performances in each index will be selected to construct the asset pool.An empirical test will be implemented to test the performance of the optimal strategy computed by Mean-CVa R model.Chapter 1 introduces the research background and significance,surveys the literatures of similarity measurement methods,Cluster Analysis algorithms and portfolio theories.And then the contents,research ideas,key difficulties and innovations are listed.Chapter 2 combs the dynamic time warping algorithm and its constrained algorithm as well as the theory of clustering analysis.Based on the concept of warping distance and cluster analysis technique,a clustering analysis algorithm via multi-indices dynamic time warping distance is proposed.Chapter 3 takes the constituent stocks of HS300 as sample pool.A few financial indices for each industry that distinguish the performances between stocks are selected to classify the stocks by cluster analysis algorithm proposed in Chapter 2.The groups of stocks with better performance in each index will be used to construct the asset pool.Chapter 4 solves the optimal portfolio strategy by Mean-CVaR model and compares the market performances in a rolling cycle mode between strategies of equal-weight portfolio,indexed-weight portfolio,and portfolios with minimum risk under the mean return of equal weights and index weights.Chapter 5 summarizes the main work,and proposes the future research based on the deficiencies of this paper.