Rule Evaluations, Attributes, and Rough Sets: Extension and a Case Study

Manually evaluating important and interesting rules generated from data is generally infeasible due to the large number of rules extracted. Different approaches such as rule interestingness measures and rule quality measures have been proposed and explored previously to extract interesting and high quality association rules and classification rules. Rough sets theory was originally presented as an approach to approximate concepts under uncertainty. In this paper, we explore rough sets based rule evaluation approaches in knowledge discovery. We demonstrate rule evaluation approaches through a real-world geriatric care data set from Dalhousie Medical School. Rough set based rule evaluation approaches can be used in a straightforward way to rank the importance of the rules. One interesting system developed along these lies in HYRIS (HYbrid Rough sets Intelligent System). We introduce HYRIS through a case study on survival analysis using the geriatric care data set.