Ratio Rule Mining with Support and Confidence Factors

Various data mining methods are being considered. This paper examines the problem of extracting ratio rules. ratio rules are linear relationships in numeric attributes applicable to understanding data, filling missing attribute values, and related issues. Existing research for ratio rules, however, does not consider a concept used in association rule mining. This prevents us from extracting a ratio rule having a strong linear relationship in part. This also prevents us from measuring objective goodness of each ratio rule. We formulated ratio rule mining in analogy to association rule mining, and introduce support and confidence concepts to ratio rules. We propose a ratio rule extraction method based on support and confidence, and show the appropriateness of our proposed method using real and synthetic data