Learning to Rank Cases with Classification Rules

An advantage of rule induction over other machine learning algorithms is the comprehensibility of the models, a requirement for many data mining applications. However, many real life machine learning applications involve the ranking of cases and classification rules are not a good representation for this. There have been numerous studies to incorporate ranking capability into decision trees, but not rules. We propose a framework for ranking with rules. The framework extends and substantially improves on the reported methods of using decision trees for ranking. It introduces three types of rule ranking methods: post analysis of rules, hybrid methods, and multiple rule set analysis. We also study the impact of rule learning bias on the ranking performance. An empirical study has been conducted to evaluate some of the above methods. The results have been compared with those reported for a decision tree with ranking capability.