Financial risk management with the help of machine training

DOI №______

Abstract

The article analyzes financial risks: the process of identifying and managing them. It is established that in order to evaluate a risk, in a static or dynamic approach, it is necessary to find its causes, conditions of occurrence and to determine the main characteristics: probability and potential loss from realization. The stages of risk management are identified, which include: risk identification, risk assessment, prediction of trends in risk situations, planning, risk control. It is established that the use of artificial intelligence in financial risk management will save time and money for firms and banks. The benefits of intelligent systems that are used in risk management before managing even a professional manager have been established. The development of the financial sector and the aspects that affect it are investigated. The development of machine learning and Big Data, as well as their use in the financial field are analyzed. The advantages and disadvantages of such use are identified. The conditions for the proper use of machine learning have been established. The models of risk assessment are investigated, their features, advantages and disadvantages are established. Improvements in machine learning algorithms are investigated. Machine learning methods are analyzed and their features are established. Machine learning has methods of varying complexity, but using simple methods is no less beneficial than using complex ones. Advantages of simple methods in comparison with complex ones are determined and substantiated. It is established that before complication of an algorithm it is necessary to make sure that it is no longer possible to increase the accuracy of its operation by only changing the parameters, so one of the main parameters, which requires careful optimization in each algorithm, is the complexity of the model. The complexity of the models is gradually increased by correcting the limitations of the simple method, which will improve accuracy and avoid retraining, as well as adding features. It is established that in order to use machine learning effectively, one must first train the system on a quality sample of data, and then make sure that it proposes the correct hypotheses in the evaluation of certain phenomena. That is, the best teacher for the machine is still the man.

Keywords: financial risks; management; machine learning; Big Data; methods of machine learning.

References
1. Глушенко С. А. Анализ функциональной полноты программных систем управления рисками // Вестник РГЭУ (РИНХ). 2012. № 2 (38). С. 53–62.
2. Шевцова Ю. В. Байесовские технологии в управлении операционными рисками // Электросвязь. 2010. № 10. С. 58–61.
3. Гусев А. Л. Сложные правила остановки непрерывного контроля // Современные проблемы науки и образования. 2013. № 3.
4. Frank A., Asuncion A. UCI Machine Learning Repository // University of California, Irvine, School of Information and Computer Sciences, 2010.