J48 list ranker based on advanced classifier decision tree induction
Marian Cristian Mihăescu, Paul Ștefan Popescu, Dumitru Dan Burdescu · International Journal of Computational Intelligence Studies · 2015
This paper presents a general purpose ranking algorithm called J48 list ranker based on a decision tree induction which exhibits extra functionality for accommodating instances and providing an efficient way to access them in a classification general setup. This computational need arises from the lack of a generic way of accessing classified instances in a ranked order. In general, classifiers are implemented only as data models and have little capability in handling and working with whole set of instances or specific instances. A preliminary study of this augmented classifier is on educational data where instances are represented by users (i.e., learners) and an initial task is to obtain a ranked list of students according with their label/class established by the classifier. This functionality is implemented by the augmented classifier since its design allows loading instances, searching this dense structure and finally accessing the obtained ranking list. Among the most important functionalities implemented by the ranker there are predecessor, successor, the ith predecessor and the ith successor.