Huskova V. Data mining methods and models for evaluating financial risks

In the work analyzed the main financial risks, the management of which is a key factor that determines the effectiveness of financial activities. Implemented an analysis of the activities of banks and other financial organizations, whose work is carried out under the influence of environmental uncertainties (market, economy, politics, etc.), a large number of variables, counterparties, individuals whose behavior can not always be predicted with acceptable accuracy. The possibility of minimizing financial risks at two levels has been considered - at the level of each individual loan and at the level of the loan portfolio as a whole. As a result of preliminary analysis, it was found that the most common methods of risk assessment for the task are linear and logistic regressions, classification trees, neural networks, Bayesian network. It is shown that to increase the efficiency of making objective decisions in the analysis of credit and market risks, it is advisable to use Bayesian networks and fuzzy neural networks, which make it possible to take into account uncertainties of probabilistic and amplitude types. These approaches are characterized by fast learning algorithms and simple interpretation of accumulated knowledge. Such features of the chosen approaches make them one of the most promising and effective tools for modeling and assessing financial risks.