Attribute related methods for improvement of ID3 Algorithm in classification of data: A review

• Main Authors: Nur Farahaina, Idris, Mohd Arfian, Ismail
• Format: Article
• Language: English
• Published: Peerj Inc. 2020
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://umpir.ump.edu.my/id/eprint/31461/1/attribute-related-methods-for-improvement-of-id3-algorithm-in-classification-of-data-a-review-5f75ea3618a5a.pdf; http://umpir.ump.edu.my/id/eprint/31461/; https://www.kansaiuniversityreports.com/
• Institution: Universiti Malaysia Pahang

Decision tree is an important method in data mining to solve the classification problems. There are several learning algorithms to implement the decision tree but the most commonly-used is ID3 algorithm. Nevertheless, there are some limitations in ID3 algorithm that can affect the performance in the classification of data. The use of information gain in the ID3 algorithm as the attribute selection criteria is not to assess the relationship between classification and the dataset’s attributes. The objective of the study being conducted is to implement the attribute related methods to solve the shortcomings of the ID3 algorithm like the tendency to select attributes with many values and also improve the performance of ID3 algorithm. The techniques of attribute related methods studied in this paper were mutual information, association function and attribute weighted. All the techniques assist the decision tree to find the most optimal attributes in each generation of the tree. Results of the reviewed techniques show that attribute selection methods capable to resolve the limitations in ID3 algorithm and increase the performance of the method. All of the reviewed techniques have their advantages and disadvantages and useful to solve the classification problems. Implementation of the techniques with ID3 algorithm is being discussed thoroughly.