Abstract—Classification is most challenging and innovative problem in data mining. Classification techniques had been focus of research since years. Logic, perception, instance and statistical concepts based classifiers are available to resolve the classification problem. This work is about the logic based classifiers known as decision tree classifiers because these use logic based algorithms to classify data on the basis of feature values. A splitting criterion on attributes is used to generate the tree. A classifier can be implemented serially or in parallel depending upon the size of data set. Some of the classifiers such as SLIQ, SPRINT, CLOUDS, BOAT and Rainforest have the capability of parallel implementation. IDE 3, CART, C4.5 and C5.0 are serial classifiers. Building phase has more importance in some classifiers to improve the scalability along with quality of the classifier. This study will provide an overview of different logic based classifiers and will compare these against our pre-defined criteria. We conclude that SLIQ and SPRINT are suitable for larger data sets whereas C4.5 and C5.0 are best suited for smaller data sets.
Index Terms—Classification, data mining, decision trees, logic based classifiers.
Aftab Ali Haider is with Center for Software Dependability, Muhammad Ali Jinnah University (MAJU), Islamabad, Pakistan (email: aftab775@yahoo.com).
Sohail Asghar is with Centre for Research in Data Engineering (CORD), Muhammad Ali Jinnah University (MAJU) Islamabad, Pakistan (email: sohail.asghar@jinnah.edu.pk).
[PDF]
Cite: Aftab Ali Haider and Sohail Asghar, "A Survey of Logic Based Classifiers,"
International Journal of Future Computer and Communication vol. 2, no. 2 pp. 126-129, 2013.