Classification of insurance risks and Bayesian approach to their analysis

Insurance is one of the most active spheres of modern business; it creates an organic part of the market economic relations, and volume of operations at the insurance market is growing rapidly. All this highlights the role and place of insurance in servicing labor and capital as well as in the process of risk management. The risk factor and the necessity to cover possible resulting damage induce the necessity of using insurance. It is shown that today exist a wide range of techniques for estimating the insurance risks. More specifically they are: expert approach, tariff based, mathematical and statistical methods, as well as theoretical descriptions of risk events directed towards establishing of existing cause/result interactions between the key variables. The classic approach is widely used that is based on regression analysis, distributions estimation theory, and algebraic calculus approaches. It was established that the techniques used today are based on Bayesian data analysis (Bayesian networks and special types of regression), that create a wide class of probabilistic models, nonlinear equations and multivariable distributions. Examples of probabilistic models construction are provided in the form of various complexity Bayesian networks to analysis of financial risks. The results, achieved with the Bayesian networks, are compared to expert estimates. It was shown that the probabilistic models provide higher forecasting quality for the risks considered. The further research will be directed towards development and application of decision support system for financial risks estimation using alternative techniques, models, and quality criteria.