Application of Regrssion Models to Forecasting Stock Return with Fundamental Variables in Tehran Stock Exchange

The stock return forecasting is of interest to many investors in financial markets. As the fundamental analysis has more powerful and reliable results than the technical one, the fundamental variables recorded in financial documents of a firm are taken into consideration. Present approach utilizes the machine learning algorithms to reach this end. In this paper, three algorithms are to be applied and fitted on data which are a nonlinear regression, tree regression, and stepwise regression. We use the financial records of some stocks in Tehran Stock Exchange (Iran), each stock has 15 feature variables as well as its corresponding return as a response variable. The results are compared to each other. Taking into consideration the synchronous validation criterion, the mean absolute error (MAE) for the validation data set is utilized.