A Rule Discovery Comparison of Stastics to Induction: KENDALL's vs. ID3 in the Dermatology and Liver Disorder DIAGNOSIS DOMAIN

Abstract

Recently, Data Mining (DM) gradually becomes an active domain as many mining techniques were developed. ID3, an induction-based method of data mining, applies information theory to get the entropy of the attribute by which the decision tree can be developed. On the other hand, Kendall’s correlation is a statistics-based method to help us to find the correlation coefficient of the attributes and the result. According to the coefficient, the decision tree can be generated. The comparison of the above methods is made to find out the way can extracts concise rules from huge datasets, which is seldom discussed in the past. In this research, two different datasets are used in comparing and the criterions are the number of rules and the depth of the generated decision tree. The result shows the ID3 performs better than Kendall’s correlation.