Large-scale Stock Selection Method Based On Lasso

Since 1870,the Chinese stock market has been expertencd many difficulties when growing.The establishment of the Shanghai Stock Exchange and Shenzhen Stock Exchange in 1990 has opened the chapter of the official development of China's emerging securities market.Since then,a number of bull and bear market replacements have also emerged.In order to obtain return of investment,investors also need to face investment risks when investing in securities.Due to the unavoidable nature of systemic risk,diversifying investment to avoid non-systematic risks has become the main way for Quantitative Investing.The Capital Asset Pricing Model(CAPM)in 1964 was proposed by American scholar William Sharpe.This model mainly weighs the relationship between risk and return.F-F factor and extended model was put forward by Fama and French to explain the stock return rate in 1993.Due to market5 inefficiency,CAPM can be redefined as ri=?i+rf+?i(rm-rf)+?i.The F-F factor and extended model is defined as ri=?i+rf+?i(rm-rf)+sismb+hihml+?i.?i represents the degree of correlation between the stock and markets,?i represents this stock's excess return caused by mispricing,si represents the scale sensitivity of the portfolio,and hi represents the value sensitivity of the portfolio.This paper selects the monthly data of the 1,173 stocks of the Shanghai and Shenzhen A shares after screening from 2010 to 2018 for a total of 108 issues per year,then use these data to perform stock fitting on the F-F factor and extended model and the CAPM model respectively.Calculating the coefficient values and alpha values of the factors of the two models separately,and then selecting the stocks with alpha>0 to invest.And gaining more return than the market.At the same time,the investment results of the F-F factor and extended model and the CAPM model' alpha stock picking strategy of the traditional asset pricing model are compared.The processing and implementation of the model results in this paper is based on the matlab language.Quantitative investments use code to implement trading ideas and optimize and improve them through historical data.This paper combines the Lasso algorithm to implement the model by setting the model period(T=6,12,24).Cycle analysis and update the portfolio through yield data for every 6 months,12 months and 24 months.For the CAPM and F-F factor and extended model,the investment strategy obtained from the 24-month-period model estimation is obviously better.The estimation result of the F_F factor and extended model is better than the CAPM model.The data analysis results show that the large-scale stock selection method based on Lasso is feasible.