Econometrics
Mojtaba Rostami; Seyed Nezamuddin Makiyan
Abstract
Stock returns forecasting is very crucial for investors, share-holders and arbiters. Different methods have been developed for this purpose. In general, there are four methods of forecasting in stock markets, which are; Technical Analysis, Fundamental Analysis, Traditional Time Series and Machine Learning. ...
Read More
Stock returns forecasting is very crucial for investors, share-holders and arbiters. Different methods have been developed for this purpose. In general, there are four methods of forecasting in stock markets, which are; Technical Analysis, Fundamental Analysis, Traditional Time Series and Machine Learning. This study is classified in the third category that is a time series prediction in which the values of a variable are predicted over time. Studies which have been done so far indicate that most of them concentrate on Neural Networks and Genetic Algorithm which are in Machine Learning class and none of them uses Bayesian approach or Exponential Smoothing and Box Jenkins techniques placed in the group of time series forecasting. This paper focuses on forecasting with time series methodology for predicting and comparing the results of the Bayesian, Exponential Smoothing and Box Jenkins methods together. In fact, the difference between this study and others is the comparison of the mentioned methods for stock return forecasting. The period of investigation was 2018- 2020, which covers daily frequency structure. Results, indicated that Bayesian method, based on the Root Mean Square Error (RMSE) criterion is the best technique for the prediction of stock returns. This is because, in addition to information derived from data, this method also uses other sources of information such as non-sample information or vague prior density as well for forecasting. Results illustrate the importance of considering the Bayesian approach in predicting stock market returns.
