Discovering the concise set of actionable patterns
ISMIS'08 Proceedings of the 17th international conference on Foundations of intelligent systems
| Hi-index | 0.00 |
The essential problem of Knowledge Discovery in Databases is to find interesting relationships, those that are meaningful in a domain. This task may be viewed as one of searching an immense space of possible actionable concepts and relations. Because the classical knowledge discovery algorithms are not able to determine if a pattern is truly actionable for the user, we present a new strategy, called extended action rules (EAR) that can be used not only for evaluating discovered patterns but also for reclassifying some objects described in the database from one state into another desired state. There are two types of EAR: object-based and rule-based. Object-based EAR is developed to generate actionable candidates directly from a decision table. Rule-based EAR is developed and implemented to assist humans acquiring actionable knowledge by automatically analyzing discovered patterns. It can be constructed from two classification rules extracted earlier from the same decision table, each defining different preferable classes. For a quicker and more effective process of EAR discovery, action forest algorithm and action tree algorithm are proposed. Support and confidence of the rules for object-based and rule-based EAR are presented to prune a large number of irrelevant, spurious, and insignificant EAR. The purpose of knowledge discovery systems is to model the real world. Data in the real world are perfect or imperfect, so system DEAR is built for extracting EAR from such data. In order to generate rule-based EAR we have to discover classification rules by using LERS-type algorithm, a new clustering-based rule discovery algorithm, and a novel CID algorithm for complete nominal, complete numerical, and incomplete data, respectively. System DEAR was tested on several public domain databases as well as on some real medical data. The results show that actionability should be considered as an objective measure rather than a subjective one. EAR are useful in many fields such as medical diagnosis and business.