A Case Based Practical Approach for Novel Data Transformation to Enhance Accuracy of Decision Tree Ensembles

Abstract

If we talk about any real world situation then we can see that all the situationsdramatically changes as the time passes. If we say this statement in some technical form,concepts changes gradually. This situation is called as Concept Drift that is the core ofany approach. Until we cannot get the accurate output for a given input parameter, theconcepts will concurrently change. To overcome this situation we use a programmableapproach that is Classifier Ensembles in which we combine several outputs and form asingle output from several. Other thing we get about is Decision Tree that is a verypopular ensemble method because Decision Trees are unstable classifiers whose outputundergoes significant changes. Data transformation is a process by which the problemrepresentation is changed and we have to manipulate the problems by using some usefultechniques. This paper firstly focuses on the popular ensembles methods, the overview ofdecision tree and uses the concept of classifier ensemble with respect to decision trees.There are mainly two problems associated with data transformation and we havedifferent approaches to resolve these problems. In this paper we consider a single noveltransformation method to resolve these problems.

Keywords: Ensemble, decision trees, data, transformation method