Theses and Dissertations
Date of Award
Master of Science (MS)
Dr. Dongchul Kim
Dr. Bin Fu
Dr. Xiang Lian
Discretization algorithm is a crucial step to not only achieve summarization of continuous attributes but also better performance in classification that requires discrete values as input. In this thesis, I propose a supervised discretization method, Global Entropy Based Greedy algorithm, which is based on the Information Entropy Minimization. Experimental results show that the proposed method outperforms state of the art methods with well-known benchmarking datasets. To further improve the proposed method, a new approach for stop criterion that is based on the change rate of entropy was also explored. From the experimental analysis, it is noticed that the threshold based on the decreasing rate of entropy could be more effective than a constant number of intervals in the classification such as C5.0.
Jonnalagadda, Sai Jyothsna, "Global Entropy Based Greedy Algorithm for discretization" (2016). Theses and Dissertations. 49.
Copyright 2016 Sai Jyothsna Jonnalagadda. All Rights Reserved.