This paper has two aims. The first is forecasting ination in Iran using Macroeconomic variables data in Iran (Ination rate, liquidity, GDP, prices of imported goods and exchange rates), and the second is comparing the performance of forecasting vector auto regression (VAR), Bayesian Vector-Autoregressive (BVAR), GARCH, time series and neural network models by which Iran's ination is forecasted. The comparison of performance of forecasting models used to forecast Iran's ination has been done based on the Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) of the models. Due to the annual values of Ination, liquidity, GDP, prices of imported goods and exchange rates at free market to estimate di erent models in this paper and compare root mean square error and Mean Absolute Percentage Error of models by which ination has been forecasted, neural network model had better performance than others models in forecasting Iran' s ination. Indeed root mean square error and Mean Absolute Percentage Error of neural network model have less value rather than root mean square error and Mean Absolute Percentage Error of other forecasting models.

Among macroeconomic variables, interest rates are the most important ones; however, the impact of interest rates on prices and returns on the stock exchange is not so clear. In the other words, the relationship between interest rates and stock prices over time is not constant. The purpose of this study is to investigate the relationship between the money market (interbank loan market) interest rates and stock market performance, as well as to investigate the predictability of price behaviour in the Tehran Stock Exchange. To determine effects as well as prediction, GARCH model is applied, also the Brock-Dechert-Scheinkman test (BDS) was used to determine the nonlinear behavior of the series for the stock market risk premium. The research findings confirmed the predictability and nonlinearity relationship between examined variables. Also, the research findings indicated that there is a meaningful relationship between the interbank loan rate with the price to income ratio of the Tehran Stock Exchange

An important factor in risk management is optimized conditional value at risk (CVaR) of the portfolio. Choose a model which calculates time depended to variance rather than the model with constant variance lead to improve data modeling. Using an appropriated method for measuring risk in financial asset returns distribution has a great utility. The main purpose of this study is implementing a hybrid procedure to calculate CVaR which, models, volatility and dynamics in clusters, and calculates CVaR value based on fat tail feature. In this case, using Extreme value theory (EVT) leads to calculate CVaR more precisely. In addition to, using some ARCH (autoregressive conditional heteroskedasticity) family models result to dynamic feature in estimating CVaR. Data were used in this study related to TEDPIX during 2001-2015. Total 2781 data were derived from Rahavard Novinand & TseCline softwares as daily. For analysis this TEDPIX data, MATLAB software and EXCELL were used. This result represented, return data distribution has fat tail. The historical simulation (HS) at 95% confidence level isn’t accurate, while the accuracy Generalized Auto-Regressive Conditional Heteroskedasticity-EVT (GARCH-EVT) model at 95% is more suitable. Using (Fractionally integrated generalized autoregressive conditional heteroskedasticity -EVT) FIGARCH-EVT method leads accurate estimates of CVaR in comparison with HS procedure. Calculating CVaR by FIGARCH-EVT-CVaR was more accurate than the GARCH-EVT-CVaR. This model has considered to both GARCH-EVT features and long memory property. The FIGARCH-EVT-CVaR model had acceptable accuracy and its exceptions are independent. In General, models which considered heteroscedastic, had an acceptable accuracy in comparing HS.