Nonparametric model test by using adaptive group LASSO method to identify the effective features in predicting the expected returns of stock portfolios

In this paper, a new nonparametric method is applied using the adaptive group LASSO for selecting features and studying which of the features provide incremental information for predicting the cross-sectional expected return. Out of many features mentioned in previous studies, the effect of 36 characteristics on the expected returns of stock portfolios in Tehran Stock Exchange (1396-1387) was investigated.The result of this study shows that only three to five features provide incremental information to predict the expected return on stock portfolios. Therefore, only the return characteristics of 2 to 1 month before the forecast, total fluctuations, beta, maximum daily returns, and the ratio of price to the highest price have the power to predict the expected return on stock portfolios. The rest of the studied features do not have the power to predict expected returns.