Resumen
The use of artificial intelligence technologies and big data analysis in risk management makes it possible to reduce the burden on experts and reduce the influence of the human factor in risk assessment. These technologies are well studied and actively used to determine the probability of known risks and assess the magnitude of the consequences when they occur, but the approach to identifying new types of risks remains poorly developed. The authors have developed an innovative approach to identifying new types of risks based on the use of artificial intelligence methods and big data analysis.The developed approach involves the identification of new types of risk in three stages: 1) identification of anomalies in the historical data array; 2) division of the identified anomalies into homogeneous clusters; 3) profiling of clusters of anomalies as potentially new types of risks, description of the characteristic features of the identified clusters. To search for anomalous observations, the authors propose to use the technology of ensembling statistical methods and machine learning methods, such as the ellipsoidal data approximation method, the local outlier level method, and the isolation forest method. To form homogeneous clusters of anomalous observations, it is proposed to use one of the cluster analysis methods selected based on the values of internal clustering quality metrics. Correlation and statistical data analysis methods are used to profile anomalous clusters as potentially new types of risks. The proposed approach, in contrast to classical risk identification technologies, makes it possible to increase the efficiency and quality of identification. The developed approach can methodologically be integrated into standard risk management processes and used in various fields of activity for automated identification of new types of risks for the purpose of their subsequent analysis and processing.