Evaluating the limits of misclassification probability. Case study of hazardous failure prediction
https://doi.org/10.21683/1729-2646-2024-24-3-18-23
Abstract
Problem definition. Many artificial intelligence systems are essentially event classification systems. They are widely used in predictive analytics. Their role as predictors of hazardous events in transportation is constantly growing. The efficient application of artificial intelligence methods largely depends on the results of misclassification. Therefore, the problem associated with the calculation or statistical evaluation of the probability of misclassification and boundary value definition is of relevance. Aim. To estimate the boundaries for the combined probability of misclassification due to two different categories of errors, i.e., misclassifications proper and statistical errors resulting from misclassification. Results. The threshold value that is used for classification was statistically evaluated. The boundary conditions for the combined probability of misclassification were established. A generalization for N-dimensional spaces and general distributions and shapes of threshold surfaces was presented. The theoretical findings were illustrated with an example of practical application.
About the Authors
H. SchäbeGermany
Hendrik Schäbe, Doctor of natural sciences, Main specialist on RAMS
Cologne
I. B. Shubinski
Russian Federation
Igor B. Shubinsky, Professor, Doctor of Engineering, Deputy Director of Integrated Research and Development Unit
Moscow
E. N. Rozenberg
Russian Federation
Efim N. Rozenberg, Professor, Doctor of Engineering, First Deputy Director General
Moscow
References
1. Leadbetter M.R., Lindgren G., Rootzén H. Extremes and related properties of random sequences and processes. Moscow: Mir; 1989.
2. Braband J., Schäbe H. On safety assessment of artificial intelligence. Dependability 2020;4:25-34.
3. Braband J., Schäbe H., The application of artificial intelligence in railway technology for safety-relevant applications – opportunities and problems. Signalling and Datacommunication 2022;(114)5:14-21.
4. Feller W.A. Introduction to Probability Theory and Its Application. J. Wiley & Sons; 1970.
5. Vapnik V.N. The nature of statistical learning theory. 2nd Edition. Springer; 1999.
6. Akaike H. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;19(6):716-723.
7. Vapnik V.N., Červonenkis A.Ja. The uniform convergence of frequencies of the appearance of events to their probabilities. Dokl. Akad. Nauk SSSR 1968;181(1968):781-783. (in Russ.)
8. Shubinsky I.B., Rozenberg E.N., Schäbe H. Innovative Methods of Ensuring the Functional Safety of Train Control Systems. Reliability Theory & Applications 2023;4(76) (18):909-920.
Review
For citations:
Schäbe H., Shubinski I.B., Rozenberg E.N. Evaluating the limits of misclassification probability. Case study of hazardous failure prediction. Dependability. 2024;24(3):18-23. (In Russ.) https://doi.org/10.21683/1729-2646-2024-24-3-18-23