Prediction of Stock Price Resilience, using Artificial Neural Networks (MLP) in companies listed on the Stock Exchange

Over the past few years, the growth and development of the capital market of countries and the introduction of new tools, mechanisms, and phenomena in it, have enhanced the importance of the capital market in the economy. In this research, the researcher aims to predict Stock Price Resilience in the Iranian Stock Exchange using the multilayer perceptron model of artificial neural networks. The present study has an applied approach that pays special attention to providing a purposeful method in designing stock price resilience forecasts. The statistical population is the companies listed on the Tehran Stock Exchange in the period 2009-2019. The dynamic artificial neural network structure design is done in the MATLAB software environment. Comparison of AR and ARMAX statistical methods and NAR and NARX networks has been used to predict the average Stock Price Resilience for the next year. The results show that in estimating Stock Price Resilience, the highest amount of R2 is present in NARX, NAR, ARMAX, AR models, respectively. This means that the best models for estimating Stock Price Resilience are listed in order. Based on the criteria of mean square error, total square error, coefficient of explanation, prediction error of the NARX model for the stock price parameter is very low, so this model has a much higher accuracy in predicting the stock price resilience than other models so Which is a lower amount of prediction error.