From data mining to rule refining. A new tool for post data mining rule optimisation

The discovery of information from data and its presentation to the user have long been the primary goals of data mining. This paper describes a new software tool, the Rule Refiner, which focuses on post data mining operations, the optimisation of rules, the visualisation of rule characteristics and their validity within domain data. The system also provides facilities for the manipulation of this information and the combination of knowledge from a variety of sources. These features allow a highly problem-specific model to be created that makes use of domain knowledge, be it discovered by algorithms and/or learnt by human experts. This can lead towards expert systems with better predictive accuracy or more understandable rule sets and sometimes even both, as data mining algorithms perform with different quality according to the domain they are applied to