The problem of modeling and forecasting possible financial loss in the form of market risk using stochastic measurements is considered. The sequence of operations directed towards risk estimation includes data preparing to model building with selected filters: exponential smoothing, optimal Kalman filter and probabilistic Bayesian filter. A short review of the possibilities for data filtering is proposed, and then some of them are selected for specific practical application to process financial data in the form of prices for selected stock instrument. After preprocessing the data is used for constructing forecasting models for the financial process itself and dynamic of its conditional variance. In the first case regression models with polynomial trend are hired, and to describe dynamic of conditional variance GARCH and EGARCH models are constructed. Further on the results of variance prediction are used for computing possible market loss hiring VaR approach. Adequacy analysis of the models constructed and back testing of risk estimates performed indicate that there is improvement of quality of the final results.

No abstract

No abstract

No abstract

No abstract