The main goal of the research is to investigate the performance of the neural network model in predicting the volatility of the index and compare it with the performance of the GARCH model. The importance of examining volatility as a measure of risk in financial markets comes from the fact that volatility in financial markets is one of the most important variables in the field of investment decisions, pricing of securities and derivatives, risk management, regulation and monetary policy. In addition, the volatility of financial markets plays an important role in the economy of countries by creating or reducing confidence and public trust. In this research, first, using the weekly data of the total index of the Tehran Stock Exchange, the fluctuation of the index is considered as the target output data, and the data of the years 2011-2013 are used for training and the data of the year 2012 are used to estimate the outputs in the short-term long-term memory network. After that, it was observed that the short-term long-term model was able to predict fluctuations with a very small error, and finally, the outputs of the Garch model and the neural network are compared using two information measures: root mean square error and mean absolute value error, which finally show The performance of neural network models is better than the GARCH model.