Computational Management Science

, Volume 11, Issue 4, pp 403–418 | Cite as

Machine-learning classifiers for imbalanced tornado data

  • Theodore B. TrafalisEmail author
  • Indra Adrianto
  • Michael B. Richman
  • S. Lakshmivarahan
Original Paper


Learning from imbalanced data, where the number of observations in one class is significantly larger than the ones in the other class, has gained considerable attention in the machine learning community. Assuming the difficulty in predicting each class is similar, most standard classifiers will tend to predict the majority class well. This study applies tornado data that are highly imbalanced, as they are rare events. The severe weather data used herein have thunderstorm circulations (mesocyclones) that produce tornadoes in approximately 6.7 % of the total number of observations. However, since tornadoes are high impact weather events, it is important to predict the minority class with high accuracy. In this study, we apply support vector machines (SVMs) and logistic regression with and without a midpoint threshold adjustment on the probabilistic outputs, random forest, and rotation forest for tornado prediction. Feature selection with SVM-recursive feature elimination was also performed to identify the most important features or variables for predicting tornadoes. The results showed that the threshold adjustment on SVMs provided better performance compared to other classifiers.


Machine learning Support vector machines Random forest  Rotation forest Logistic regression Tornado detection 

Mathematics Subject Classification (2010)

62H30 68Q32 62J86 



Funding for this research was provided under the National Science Foundation Grants AGS0831359 and EIA-0205628.


  1. Bi J, Bennett KP, Embrechts M, Breneman CM, Song M (2003) Dimensionality reduction via sparse support vector machines. J Mach Learn Res 3:1229–1243Google Scholar
  2. Bluestein HB (1993) Synoptic-dynamic meteorology in midlatitudes: volume II: observations and theory of weather systems. Oxford University Press, New YorkGoogle Scholar
  3. Boser BE, Guyon IM, Vapnik VN (1992) A training algorithm for optimal margin classifiers. Paper presented at the Proceedings of the fifth annual workshop on computational learning theory, Pittsburgh, Pennsylvania, USGoogle Scholar
  4. Breiman L (2001) Random Forests. Mach Learn 45(1):5–32. doi: 10.1023/a:1010933404324 CrossRefGoogle Scholar
  5. Cárdenas AA, Baras JS (2006) B-ROC curves for the assessment of classifiers over imbalanced data sets. In: Proceedings of the 21st national conference on artificial intelligence (AAAI 06), Boston, Massachusetts, July 16–20, 2006Google Scholar
  6. Donaldson RJ, Dyer RM, Krauss MJ (1975) An objective evaluator of techniques for predicting severe weather events. In: Ninth conference on severe local storms, Norman, OK, 1975. American Meteorological Society, pp 321–326Google Scholar
  7. Drummond C, Holte RC (2003) C4.5, class imbalance, and cost sensitivity: why under-sampling beats over-sampling. In: Workshop on learning from imbalanced data sets II, ICML, Washington, DC, 2003Google Scholar
  8. Efron B, Tibshirani R (1993) An introduction to the bootstrap. In: Monographs on statistics and applied probability, vol 57. Chapman & Hall, New YorkGoogle Scholar
  9. Guyon I, Weston J, Barnhill S, Vapnik V (2002) Gene selection for cancer classification using support vector machines. Mach Learn 46(1–3):389–422. doi: 10.1023/a:1012487302797 CrossRefGoogle Scholar
  10. Hall M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten IH (2009) The WEKA data mining software: an update. SIGKDD Explor Newsl 11(1):10–18. doi: 10.1145/1656274.1656278 CrossRefGoogle Scholar
  11. Hand DJ, Mannila H, Smyth P (2001) Principles of data mining. In: Adaptive computation and machine learning. MIT Press, CambridgeGoogle Scholar
  12. Heidke P (1926) Berechnung des erfolges und der gute der windstarkvorhersagen im sturmwarnungsdienst. Geografiska Annaler 8:301–349CrossRefGoogle Scholar
  13. Kubat M, Matwin S (1997) Addressing the curse of imbalanced training sets: one-sided selection. In: Proceedings of 14th international conference on machine learning, 1997. Morgan Kaufmann, Los Altos, pp 179–186. citeulike-article-id:2526066Google Scholar
  14. Lakshmanan V, Stumpf G, Witt A (2005) A neural network for detecting and diagnosing tornadic circulations using the mesocyclone detection and near storm environment algorithms. In: 21st international conference on information processing systems, San Diego, CA, 2005. p J5.2Google Scholar
  15. Marzban C, Stumpf GJ (1996) A neural network for tornado prediction based on Doppler radar-derived attributes. J Appl Meteorol 35(5):617–626CrossRefGoogle Scholar
  16. McGill R, Tukey JW, Larsen WA (1978) Variations of box plots. Am Stat 32(1):12–16. doi: 10.2307/2683468 Google Scholar
  17. Platt J (1999) Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. In: Smola A, PB, Schölkopf B, Schuurmans D (ed) Advances in large margin classifiers. pp 61–74. citeulike-article-id:3115812Google Scholar
  18. Provost F, Fawcett T (2001) Robust classification for imprecise environments. Mach Learn 42(3):203–231CrossRefGoogle Scholar
  19. Provost FJ, Fawcett T, Kohavi R (1998) The case against accuracy estimation for comparing induction algorithms. Paper presented at the proceedings of the fifteenth international conference on machine learningGoogle Scholar
  20. Richman MB (1986) Rotation of principal components. J Climatol 6(3):293–335CrossRefGoogle Scholar
  21. Rodriguez JJ, Kuncheva LI, Alonso CJ (2006) Rotation forest: a new classifier ensemble method. IEEE Trans Pattern Anal Mach Intell 28(10):1619–1630. doi: 10.1109/TPAMI.2006.211 CrossRefGoogle Scholar
  22. Roebber PJ (2009) Visualizing multiple measures of forecast quality. Weather Forecast 24:601–608CrossRefGoogle Scholar
  23. Stumpf GJ, Witt A, Mitchell ED, Spencer PL, Johnson JT, Eilts MD, Thomas KW, Burgess DW (1998) The national severe storms laboratory mesocyclone detection algorithm for the WSR-88D. Weather Forecast 13(2):304–326CrossRefGoogle Scholar
  24. Trafalis TB, Ince H, Richman MB (2003) Tornado detection with support vector machines. Paper presented at the proceedings of the (2003) international conference on computational science. Melbourne, AustraliaGoogle Scholar
  25. Trafalis TB, Santosa B, Richman MB (2004) Bayesian neural networks for tornado detection. WSEAS Trans Syst 3:3211–3216Google Scholar
  26. Trafalis TB, Santosa B, Richman MB (2005) Learning networks for tornado forecasting: a Bayesian perspective. WIT Trans Inf Commun Technol 35:5–14Google Scholar
  27. Vapnik VN (1998) Statistical learning theory. In: Adaptive and learning systems for signal processing, communications, and control. Wiley, New YorkGoogle Scholar
  28. Wilks D (1995) Statistical methods in atmospheric sciences. Academic Press, San DiegoGoogle Scholar
  29. Yang JH, Honavar V (1998) Feature subset selection using a genetic algorithm. IEEE Intell Syst App 13(2):44–49. doi: 10.1109/5254.671091 CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Theodore B. Trafalis
    • 1
    Email author
  • Indra Adrianto
    • 1
  • Michael B. Richman
    • 2
  • S. Lakshmivarahan
    • 3
  1. 1.School of Industrial and Systems EngineeringThe University of OklahomaNormanUSA
  2. 2.School of MeteorologyThe University of OklahomaNormanUSA
  3. 3.School of Computer ScienceThe University of OklahomaNormanUSA

Personalised recommendations