Research On Stock Returns Based On Functional Data And Cluster Analysis

The stock market is a very important part of the financial market.The operation of the stock market is based on the fluctuation of the stock price,which directly affects the investors' investment interests.When investors invest in stocks,they usually make decisions according to the macroeconomic indicators and the company's financial indicators.In stock market trading,the data of stock price is changing at any time,and the change of stock price often affects the behavior of investors.Compared with other financial indicators,the change of stock price is real-time and rapid.Some investors may pay more attention to the real-time data than the updated financial data every quarter.Since the stock price data contains huge information,will it be more accurate and effective to use these data to study the stock return? At present,most of the stock investment strategies constructed by securities research institutes are based on the industry classification,which include the CSRC industry classification,Shenwan industry classification,CITIC industry classification,etc.This kind of classification method is effective in many occasions.This paper gives a new stock classification method which is based on stock price data,and studies whether the portfolio constructed in this way will have a higher yield.This paper selects the trading day data from January 1,2018 to December 31,2018,and selects the daily return rate,dynamic P/E ratio,net assets per share,Shanghai Bank's opening rate and the per minute stock price data of each share in the morning.Because the stock price data is a function type data on time t,in Chapter 4,the regression model of daily return and other independent variables is fitted by using the function type partial linear model.The coefficient ?(t)which can describe the fluctuation trend of stock price is obtained.Then,based on this coefficient curve,curve clustering algorithm is used to classify the sample stocks.All stocks are divided into five different categories according to the fluctuation trend.According to this classification result and Shenwan industry classification standard,the paper constructs the timing investment strategy and calculates the total investment return rate respectively.In Chapter 5,the investment strategy is simulated 10000 times.The empirical results show that the average value of total return of stock portfolio based on the trend clustering of stock price volatility is higher than that based on the industry classification of Shenwan,and the standard deviation is slightly lower than that based on Shenwan,and both of them are higher than the total return of HS300 index in the same investment period.