Financial risk management via multi model inference grid applications

A pivotal component of modern risk management is the employment of dynamic statistical models to forecast risk and return of financial assets. These forecasts are the fundamental ingredient on the basis of which the decisions of financial agents are made. As the degree of complexity of these models can be quite high, in recent years the software industry has developed automated systems which provide periodic forecasts from some “benchmark” models which are ready to use for financial practitioners. Although very popular in financial consulting firms, these products are not useful for more advanced analysis and research as a consequence of the static nature of the modelling approach: forecasts are usually produced from one model which is claimed to be optimal on the basis of often not very sound statistical and economic considerations. A much more appealing software infrastructure for financial researchers and practitioners would be one capable of providing an interactive modelling approach. By interactive modelling we mean a software environment which allows users to define which models to use, share the empirical evidence drawn from the different models of the community of users and combine different forecasts or choose the optimal one in order to achieve superior predictive ability. Recent developments in statistical Multi Model Inference (MMI) and GRID computing make this type of applications feasible. This work presents the implementation of an MMI-GRID application which aims at reaching these goals in the context of volatility analysis and forecasting.