Locality-aware ratio rule mining

This paper examines a Ratio Rule mining problem based on support and confidence factors. Ratio Rules represent linear relationships included in some numeric attributes. They are applicable to understanding data, filling missing attribute values, and related issues. In our previous work, we have formulated Ratio Rules in analogy with association rule mining, and proposed a mining algorithm. However these Ratio Rules consider "Global" relationships. In this paper we formulate the Local Ratio Rules, which introduces locality concept, and propose a method to extract them. We show appropriateness of our proposed method using synthetic and real data.