Comparison and analysis of stock futures return response to non-systematic risk torque measurement models(comparative prediction with neural network

The mean and variance of stock returns alone are not sufficient to describe the distribution of returns. , paying attention to higher torques such as skewness and kurtosisas a risk index instead of variance leads to more accurate results. Therefore, due to study contradictions, the study of stock returns reaction to models for measuring non-systematic risk moments was significant. For this purpose, 152 companies were selected( In the realm of time between 2014to 2021 ) as a statistical sample from the companies listed on the Tehran Stock Exchange based on the systematic removal method (CAPM & FF3). Also, by determining the superior regression model, the power of the superior regression model was compared with the neural network model. results showed that by increasing the unsystematic risk torques calculated with the capital asset pricing model and the Fama and French tree-factor model reduce future stock returns; These results can be justified in line with the concepts of capital market efficiency theory.Other results indicate that the difference in expected non-systematic risk calculated with CAPM and FF3 models is significant. Therefore, the strength of the expected non-systematic risk torques calculated by the capital asset pricing model is less than the Fama and French three-factor model. By comparing the neural network model coefficient and linear regression, it can be said that the neural network-based model has a better performance in predicting future stock returns based on non-systematic risk moments than linear regression.