Logical and Probabilistic Models of Banking and Financial Risks with Groups of Incompatible Events
17 Pages Posted: 10 May 2001
Date Written: undated
Abstract
We are considering the structural complex systems where the risk is usual and mass phenomenon and there is statistics in enough amount about risk objects. The examples of such systems take place in banks, business, insurance, quality, monitoring systems. The tasks of classification of objects on "good" and "bad" are solved by means of discriminant analysis methods (DA), neural networks (NN) and artificial intelligence (AI). There is a immense number of publications on these methods of classification. The task of classification on "good" and "bad" objects by DA methods is in construction of simple approximating mathematical model. Usually the flatness or convex surface are such simple models in n-dimensional space of classified signs which are devoting the objects on two classes. The relative average error in classification Em can achieve 0.15-0.3 (of general number of objects). The algorithm of construction of strict separating plane is unknown and cannot be realized in practice.
Suggested Citation: Suggested Citation