Stock Return Forecasting Using the Bayesian Model Approach in Tehran Securities Exchange

In the present study, stock returns were predicted using the Bayesian model approach in the Tehran Stock Exchange. Therefore, the research hypothesis based on the Bayesian method has higher accuracy in predicting returns than Autoregressive models was developed and tested. In order to test the research hypothesis, information related to the index of 30 selected industries in the Tehran Stock Exchange during the period from 2017-03-25 to 2020-08-24 was used. The index return was predicted based on two methods for 30 out-of-sample data. First, autoregressive models were fitted on the returns of each index and then the next 30 days of returns were predicted based on these models. Then, after identifying the optimal model lags through the Bayesian model averaging method, autoregressive models were fitted with the optimal lags and the next 30 days predictions were obtained under this method. In order to compare the accuracy of the methods in predicting the return, RMSE and MAE criteria were used and the values of these error criteria were compared using Wilcoxon nonparametric pairwise comparison tests. The results showed that the Bayesian method leads to an increase in the accuracy of model prediction in out of sample data.