An improved Id3 algorithm for medical data classification. (January 2018)

• Record Type: Journal Article
• Authors: Yang, Shuo; Guo, Jing-Zhi; Jin, Jun-Wei
• Is Part Of: Computers & electrical engineering. Volume 65(2018)
• Page Start: 474
• Page End: 487
• Publication Date: 2018-01
• Subjects: Id3 algorithm -- Decision tree -- Balance function -- Numeric attribute discretization -- Rule representation of classifier model; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2017.08.005
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: An improved Id3 algorithm is proposed for accurate and reliable disease prediction. The multi-value bias problem of the Id3 algorithm is mathematically proved. The reasonability of the new balance function is proved by theory and experiments. The best partition number for numeric attribute discretization is proved. The novel rule-based classification model reduces average memory cost. Abstract: Data mining techniques play an important role in clinical decision making, which provides physicians with accurate, reliable and quick predictions through building different models. This paper presents an improved classification approach for the prediction of diseases based on the classical Iterative Dichotomiser 3 (Id3) algorithm. The improved Id3 algorithm overcomes multi-value bias problem when selecting test/split attributes, solves the issue of numeric attribute discretization and stores the classifier model in the form of rules by using a heuristic strategy for easy understanding and memory savings. Experiment results show that the improved Id3 algorithm is superior to the current four classification algorithms (J48, Decision Stump, Random Tree and classical Id3) in terms of accuracy, stability and minor error rate.