Optimization of classification trees: strategy and algorithm improvement Original Research Article

We present a new version of the program MedTree which is of use to calculate classification trees beyond the quality of trees which are based on direct evaluation of a splitting criterion. MedTree calculates a large number of possible segments of trees and recursively selects the best of these parts to form an ‘optimal’ tree which requires the discussion of the definition of ‘optimal’. The characteristics of the improved version are as follows: analysis of incomplete data, higher performance, reduction of field dimensions, higher portability, rejection of irrelevant parts of trees in an early stage, additional coupling scheme to produce either very short or longer trees with better quality by coupling the selection rule to the vicinity of an incomplete tree to the conditions of closing a tree and a interactive mode to view and adjust the selection and closing rules. Examples in order to demonstrate the improvements are given.