Increasing the reliability of risk assessments in the monitoring processes described by general distributions

 Y.E. Shishkin1,2, A.V. Skatkov1

1 Federal State Educational Institution of Higher Education «Sevastopol State University», Russian Federation, Sevastopol, Universitetskaya St., 33

2 Institute of Natural and Technical Systems, Russian Federation, Sevastopol, Lenin St., 28

E-mail: iurii.e.shishkin@gmail.com

DOI: 10.33075/2220-5861-2019-1-41-47

UDC 004.9   

Abstract:

   The article analyzes the influence of the decision-making strategy of threshold values choice on the risk of the first and second kind errors in identifying anomalies caused by a shift in observation and an increase in dispersion. We consider the processes of monitoring the parameters of environment objects and processes, which in practice are described by general form distributions. In the course of a purposeful experiment, on a simulator stand, the anomalies of two classes were considered: a shift in observation and an increase in dispersion, the sensitivity of the statistical detection method was also estimated. The proposed approach makes it possible to determine how much the reliability of the determining anomalies function changes when the distribution law of the input parameter changes from normal to general distribution, in particular, exponential, Weibull and gamma distributions are considered.

   A mathematical model has been developed and numerical experiments have been carried out which allow one to take into account errors of the first and second kind, with different values of shear and dispersion parameters. The application of the developed model allows obtaining updated estimates of the confidence interval for various distribution laws, which increases the reliability of the anomaly identification system, thus a significant improvement of the quality of decision making is achieved by taking measures to compensate for these errors, resulting in increased decision making reliability.

   Practical recommendations on the choice of the decision support system thresholds for identification of anomalies, taking into account the distribution of the observed variable are formulated.

Keywords: monitoring, mathematical modeling, anomaly detection, clustering, critical systems, data mining, general distribution.

Full text in PDF (RUS)

LIST OF REFERENCES

  1. Anomalies of the upper water column in the Mediterranean Sea / I. Rivetti, F. Boero, S. Fraschetti [et al.] // Global and Planetary Change. 2017. Vol. 151. P. 68–79. DOI: 10.1016/j.gloplacha.2016.03.001
  2. Shishkin Yu. E., Skatkov A.V. Method for detecting anomalies in an interactive mode in observations scalar fields gradients // Monitoring systems of environment. Sevastopol: INTS. 2018. issue 12 (32). P. 30–37.
  3. Lepikhin A. P., Voznyak A. A. Statistical distribution functions of hydrochemical indicators of water quality of surface water objects // Water management in Russia: problems, technologies, management. 2012. No. 4. P. 21-32.
  4. Zinchenko V., Falko V. Mathematical model and solution method elaboration for the task of ecological risk predictive assessment // Environmental safety. 2013. № 2 (16). P. 36–39.
  5. A.V. Skatkov, Y.E. Shishkin. Anomaly identification model in the observation field using parametric monitoring systems // Monitoring systems of environment. Sevastopol: INTS. 2017. № 10 (30). P. 48–53.
  6. A PDCA-based approach to Environmental Value Stream Mapping (E-VSM) / J.A. Garza-Reyes, J.T. Romero, K. Govindan [et al.] // Journal of Cleaner Production. 2018. Vol. 180. P. 335–348. DOI: 10.1016/j.jclepro.2018.01.121.
  7. Y.E. Shishkin, A.V. Skatkov Mobile cloud micro services actor model of monitoring // Monitoring systems of environment. Sevastopol: INTS. 2018. № 14 (34). P. 56–62. DOI: 10.33075/2220-5861-2018-4-56-62.
  8. Y.E. Shishkin, A.N. Grekov. The concept of automated environmental monitoring intellectual system of the Sevastopol Bay based on compact autonomous robots // Monitoring systems of environment. Sevastopol: INTS. 2018. № 14 (34). P. 63–69. DOI: 10.33075/2220-5861-2018-4-63–69.
  9. Vychuzhanin P., Hvatov A., Kalyuzhnaya A.V. Anomalies Detection in Metocean Simulation Results Using Convolutional Neural Networks // Procedia Computer Science. 2018. Vol. 136. P. 321–330. DOI: 10.1016/j.procs.2018.08.282
  10. Comparison of atmospheric particle concentration measurements using different optical detectors: Potentiality and limits for air quality applications / A. Dinoi, A.Donateo, F.Belosi [et al.] // Measurement. 2017. Vol. 106. P. 274–282. DOI: 10.1016/j.measurement.2016.02.019
  11. Bryukhovetsky A. A., Skatkov A.V., Shishkin Yu. E. Modeling of anomaly detection processes in complex structured monitoring data // Monitoring systems of environment. Sevastopol: INTS. 2017. No. 9 (29). P. 45-49.
  12. Harman G. Modern factor analysis. M.: Statistics. 1972. P. 489.
  13. Semenov V. A. Probability theory and mathematical statistics: Textbook. Standard of the third generation. SPb.: Piter. 2013. P.192.
  14. Shishkin Y.E. Big Data visualization in decision making // Science in Progress: тез. Всерос. науч.-практ. конф. магистрантов и аспирантов. Новосибирск, 20 октября 2016 г. Новосибирск: НГТУ, 2016. C. 203–205.

Loading