Research On Stock Portfolio Strategy Based On Functional Data Clustering Analysis

Since the establishment of the Shanghai Stock Exchange and the Shenzhen Stock Exchange in 1990,China’s stock exchange market has developed rapidly.And with the further integration of the global capital market,the risks have intensified.Therefore,to establish an effective portfolio has become the issue of concern to individual and institutional investors.Meanwhile,the amount of financial data is increasing geometrically,not only with complex fluctuations,but also with high interference.Given that the amount of today’s data is keep expanding,Functional Data Analysis,compared with traditional analysis methods,is more capable in handling data analysis job.In terms of processing massive amounts of complex data,Functional Data Analysis can achieve pattern mining of infinite-dimensional data and dig out more useful information.The article takes the 43 constituent stocks in China’s Shanghai 50 Index as the research object.Collects data and calculates the daily rate of return.Then it performs data preprocessing and function data fitting.After converting to function data,it processed descriptive analysis and functional principal component analysis of functional data.Then the results will be used to cluster 43 stocks based on the K-means clustering method.Finally,clustered results will be used to select portfolios with different strategies to verify the effect of the portfolio,aiming to provide investors with a choice of portfolio strategies.The research results in this paper show that:(1)The Functional Data Analysis can well excavate deep information of target stocks,use it to perform clustering,which can distinguish stocks with different return trend.(2)When using Functional Data Analysis methods for data reconstruction and fitting,the smoothing parameters λ comprehensively take of overfitting and over smoothing into consideration,hence its effect is better than taking the minimum value of GCV directly.(3)Using the strategy based on clustering results could achieve significantly higher investment returns than other strategies.As Functional Data Cluster Analysis can be used to cluster stocks with different gains,the shares of different trend categories can be selected and combined based on the clustering results,and the portfolio with the highest return on investment can be further selected.It is proved that the Functional Data Clustering Analysis method has particular practical significance in stock clustering and portfolio strategy selection,and it is of reference significance for investors to choose a portfolio.The innovation of this paper lies in: it adopts Functional Data Analysis methods on the research of stock portfolio strategies,which makes up for the shortcomings of traditional analysis methods,who can only analyze scattered points and the deficiency of information mining.While,when comparing the returns of various portfolio strategies,Functional Data Analysis methods take into account the risk factors,the riskadjusted return,that is,the Sharpe ratio,as the measurement index.