Hierarchical Classifiers Based on Kolmogorov-Smirnov Distance

Student: Sergey Ivanushkin

Project Goals

   The objective of this project is

a.   To study the existing hierarchical classifiers presented by binary decision trees in which each decision involves the comparison of a single feature to a threshold.

b.   To develop the mathematical model of binary decision tree classifiers.

c.    To develop algorithm for partitioning of a feature space based on Kolmogorov-Smirnov test which requires the calculations of K-S distances and threshold coefficients.

d.   To design a hierarchical pattern classifier which includes friendly GUI, the software implementation of the above algorithm, all necessary visualization and output files in appropriate format.

e.    Perform experiments on real data sets, test and evaluate the results by using the Kolmogorov-Smirnov measure of confidence.  

f.    Analysis of experimental results.