چکیده:
Using methods to predict future states has always been a primary concern for thinkers in various sciences. In this regard, naturally, some methods possess appropriate durability and applicability, having the minimum possible error in prediction. Based on this, for many years, mathematical methods—such as simple moving average, weighted moving average, double moving average, regression, and similar ones—were the only patterns decisively approved and used; however, they also had flaws at various times. With the creation of artificial intelligence methods, such as neural networks, especially in cases where a suitable mathematical relationship between data and independent and dependent variables could not be formed, much hope was created. This hope continued to the extent that it was even considered a replacement for mathematical methods. In this article, by examining the performance of neural network models and the regression model in the field of data related to stock price prediction, the measurement of prediction errors for these two methods has been addressed. The research method employed in this article is the evaluation method.
خلاصه ماشینی:
Stock Price Prediction Modeling Using Neural Networks and Comparing It with Mathematical Prediction Methods Abbas Toloie Ashlaghi* Shadi Haghdoust** \The use of methods to predict future conditions has always been a primary concern for thinkers in various sciences.
In this article, by examining the performance of neural network models and the regression model in the field of data related to stock price prediction, the prediction errors of these two methods have been addressed.
Keywords: Prediction, stocks, stock price, neural network method, regression model Introduction Investment and capital accumulation have played a significant role in the economic transformation of a country.
In this research, the variable of the daily closing stock price was used, and in the end, they reached the conclusion that index prediction by the neural network for two stock exchanges provides a more appropriate and acceptable response compared to the other two stock exchanges.
In fact, the aim of this article is to present an intelligent model for predicting stock prices in the Tehran Stock Exchange and comparing it with methods based on mathematical modeling.
Considering the opinions of experts who had appropriate experiences in various fields of neural network modeling and given the characteristics of the Tehran Stock Exchange, the number of training data and test data were selected as follows: (1).
The test data - which consists of five data points - were tested with both methods, and the results are as follows: (Refer to the page image) As can be seen, the Mean Squared Error of the neural network was higher compared to the regression model.
(
, 
, 
)
