Comparison of Linear Regression Models Ordinary Lasso, Adaptive Group Lasso and Ordinary Least Squares Models in Selecting Effective Characteristics To Predict the Expected Return

In this study, for the selection of the characteristics of the company that provides the incremental information to investors and financial analysts, the linear models are adapted by the ordinary lasso method (tibshirani, 1996), adaptive group lasso (zu, 2006) and the least squares method (ols). the main objective of this research is to determine which method can predict the expected return on stock portfolios in the shortest time and using the least effective features. the research sample is1340observations, including 134companies listed in tehran stock exchange, and the research variables from the financial statements of the companies and the stock market reports between 2008and 2018. the results of this study show that by employing the least squares regression method, 7 characteristics, the typical 5- characteristics lasso method and in the adaptive group lasso method, only 4characteristics, contain incremental information to predict the expected returns of stock portfolios. in the second place, by applying the adaptive group lasso regression method, one can achieve the same results with using the least characteristics.