the evaluation of the capability of the regression & neural network models in predicting future cash flows

Cash flow and profit are two important indicators for measuring the performance of a business unit. the future prediction was always a necessity in everyday life, and one of the subjects in which “the prediction” has a great importance is economical and financial problems. the purpose of the present study is to predict future cash flows using regression and neural network models. sub – separated variables of the accruals and operational cash flows were used to investigate this prediction. for this purpose, data of 137 accepted stock exchange companies in tehran during 2009 to 2017 has been studied. in this study, eviews9 software for regression model and matlab13 software for multi-layer artificial neural networks (mann) with error back propagation algorithm were used to test the hypotheses. the findings of the research show that both regression and neural network models within proposed variables in the present study have the capability of predicting future cash flows. also, results of neural network models' processes show that a structure with 16 hidden neurons is the best model to predict future cash flows and this proposal neural network model compared with regression model in predicting future cash flows has a better and accurate function. furthermore, in this study, it was noticed that accruals of assets compared with debt accrual and variables of operating cash flows with accrual components were more predictive for future cash flows