The purpose of this study is to compare the accuracy of predicting market risk calculation methods of value at risk with the relevance of the artificial intelligence approach. the increasing development of financial markets has revealed the importance of estimating the well-known measure of market risk, risk value more than before. Value at Risk (VaR) is a statistical measure that calculates and quantifies the maximum expected loss from holding an asset or portfolio over a period of time with a certain probability (known confidence level) and is one of the most important market risk criteria that is widely used to manage financial risk by financial regulators and portfolio managers. Macro-level risks have pervasive effects and can have negative effects on the entire financial market. Recognition of the interdependence and mutual relevance of companies and the development of risk factors that predict an increase of dependence on the sequence returns to the company during the crisis is of utmost importance. The existence of such methods provides a powerful tool for decision-makers to increase future financial stability. After reducing the fluctuations of Bootstrap, Historical, and Variance covariance with using wavelet transformation for model training and forecasting, the method is used every 15 consecutive days as input (same as the independent variable) in RVM model and 16th day as the dependent variable was considered as the model output and to evaluate the models, two evaluation criteria called Mean Square Error (MSE) and Mean Absolute Error (MAE) have been used. The relevance vector machine algorithm was used to predict variables. The RVM algorithm is a nonlinear model and causes the algorithm to be nonlinear by transferring data from the input space to the property space. The reason for checking with the MAE error is that this error represents the mean of the absolute value and is more comprehensible for us than the MSE, which is the Mean Squared Error. The results of testing the hypotheses and fitting the relevant artificial intelligence algorithm showed that the artificial intelligence algorithm is an efficient method for predicting daily value-at-risk methods. Also, in the Iranian capital market, risk-value forecasting is done with the semi-parametric bootstrap method with higher power and is recommended for use. Parametric methods (variance-covariance) and historical simulation are in the next ranks. Studies on value at risk have been limited to one industry or by portfolio definition and all listed companies have not been examined.
In this study, we tried to calculate the market risk of all companies listed on the stock exchange with the value-at-risk approach under 3 important and widely used models: variance-covariance, historical simulation, bootstrap simulation and measure their efficiency using an artificial intelligence algorithm. In a way, previous researches have a less statistical population and do not measure the efficiency of models in practice.