Application Of Penalized Regression And Support Vector Regression Machine And Their Combinations In Fundamental Quantitative Investment

Fundamental quantitative investment is an investment method that has attracted more and more attention.Its nature is to use quantitative methods to study the relationship between stock fundamentals and yields.This paper selects two representative machine learning models,the penalized regression model and the support vector regression machine model,to study their application in fundamental quantitative investment.In addition to comparing their performances,this article combines them to study the performance of the their combination.This paper selects 17 factors as independent variables to predict stock return rate,and selects data from 2010 to 2020.This paper takes quarter as a unit,each quarter as a period,takes the data of every four years as a training set to train the model,uses the factors of current period to predict the stock return rate in the next period,and constructs the portfolio according to the predicted value of stock return rate.This paper uses two types of combinations.The first is a type of parallel combination,that is,the prediction results of the two models are summed according to a certain ratio to obtain a new prediction result.The second is a type of series combination,which uses the feature screening ability of Lasso regression and Elastic-net regression in penalized regression to perform feature screening on the data,and then use the support vector regression machine to predict the rate of return.By comparing the performances of the corresponding investment portfolio,this paper finds that the performance of the support vector regression machine is better than that of the penalized regression.Their series combination improves the return performance of the investment portfolio.The performances of these models are better than CSI300.This paper verifies the effectiveness of these models on fundamental quantitative investment.