Abstract
The classification of large dimensional data sets arising from the merging of remote sensing data with more traditional forms of ancillary data causes a significant computational problem. Decision, tree classification is a popular approach to the problem and an efficient form for representing a decision process in hierarchical pattern recognition systems. They are characterized by the property that samples are subjected to a sequence of decision rules before they assigned to a unique class.
The scope of development presented here is limited to a certain class of decision trees: binary decision trees in which each decision involves the comparison of a single feature to a threshold. Since any combination of features can be expressed as a newly defined single feature, the restriction to binary trees can be made without loss of generality.
The algorithm for partitioning of a feature space developed in this paper is based on the Kolmogorov-Smirnov test (K-S test), which requires the calculation of K-S distance and the threshold coefficients of the tree nodes.