Association Rule-based Classifier Using Artificial Missing Values

In this paper, we propose a rule-based classification method that uses artificial missing values to improve the effectiveness and precision of medical data analysis. We apply artificial missing values to avoid the sharp boundary problem encountered when discretizing continuous variables. In discretization, we treat attribute values near the boundary as missing values. We evaluated the performance of the proposed artificial missing value-based classification method and our experimental results using medical data show this method to be effective for classification. The proposed method can reduce the number of rules required to build a classifier. It may also be able to control the relation between a false positive and true positive in rule-based classifiers.