Cmer: Classification Based on Multiple Excellent Rules

One of the traditional rule-based classification tasks is to build a set of high quality classification rules for prediction. Traditional rule-based classification approaches can achieve high efficiency. However, some traditional rule-based classification methods usually generate few rules. They may miss some high quality rules, especially when the training data set is small. Therefore their accuracy may not be high in some data sets. In this paper, we propose a new classification approach called CMER (classification based on multiple excellent rules). CMER is distinguished from other traditional rule-based classification methods in three aspects. First, CMER constructs a candidate set and a seed set. Second, CMER connects the seed set with the candidate set to produce more classification rules at a time. Third, CMER uses the minimum support and foil gain to update the seed set. As a result, CMER generates more excellent classification rules, especially when the training data set is small. Our experimental results show that CMER gets higher accuracy than some traditional rule-based classification methods.