Dividend is one of the important financial factors which is of interest to managers, investors and financial analysts, and is often used for decision-makings on investment, profitability assessment and profit-related risk assessment, and to make judgments about the stock prices. Therefore, its prediction is important for both managers and stakeholders. The aim of the present study was to provide a model for predicting the dividend using the multilayer perceptron neural network (MLP). To do that, 43 companies from the metal industries of Tehran Stock Exchange were considered as the research population, and 301 years-companies for the period of 2009 to 2015 were selected as the research sample. The results of the research indicated that the perceptron neural network had a high potential for dividend prediction. The mean squared error of this network was about 0.4. Moreover, considering the architecture chosen to develop and train the network and the results of the testing data, the independent variables of the study were able to explain 80 percent of the changes in the dividend. As a result, the proposed model in the research hypothesis was accepted.
