DEPTH BASED CLASSIFICATION METHOD ON THE BASIS OF A SCALABLE MAHALANOBIS DISTANCE FOR SETS WITH UNEQUAL PRIOR PROBABILITIES

Alexander Galkin A., Kiev National Taras Shevchenko University, Kiev

pages 139-147

DOI: 10.1615/JAutomatInfScien.v48.i2.70

A nonparametric depth based method of classification is proposed for the case when data sets have unequal prior probabilities and do not belong to a common family of elliptical distributions. The multipurpose depth based classifier is developed that is independent of deviations in the location, shift model or violation of monotonous character of density functions. Scalable Mahalanobis distance is estimated at each point using the residual passage method.

  1. Pollard D., Convergence of stochastic processes, Springer-Verlag, New York, 1984.
  2. Zuo Y., Serfling R., Structural properties and convergence results for contours of sample statistical depth functions, The Annals of Statistics, 2000, No. 28, 484–497.
  3. Holmes C.C., Adams N.M., A probabilistic nearest neighbor method for statistical pattern recognition, Journal of the Royal Statistical Society, 2002, No. 64, 295–306.
  4. Godtliebsen F., Marron J.S., Chaudhuri P., Significance in scale space for bivariate density estimation, Journal of Computational and Graphical Statistics, 2002, No. 11, 1–22.
  5. Hall P., Large sample optimality of least squares cross validations in density estimation, The Annals of Statistics, 1983, No. 11, 1156–1174.
  6. Mosler K., Multivariate dispersions, central regions and depth, Springer-Verlag, New York, 2002.
  7. Silverman B.W., Density estimation for statistics and data analysis, Chapman and Hall, London, 1986.
  8. Lachenbruch P., Mickey M., Estimation of error rates in discriminant analysis, Technometrics, 1968, No. 10, 1–11.
  9.  Hoeffding W., Probability inequalities for sums of bounded random variables, Journal of the American Statistical Association, 1963, No. 58, 14–27.