ON POST-RULE MINING OF INDUCTIVE RULES USING A QUERY OPERATOR

摘要

Most Data Mining applications produce rules that are not used to query the task relevant data but only evaluate their interestingness. Some applications focus on specifying a data mining language for the mining task and do not combine rule querying and database retrieval. Most other applications employ rule querying as a post-data mining operation which merely retrieves the mined rules from a rule base for query optimisation purposes. An inductive technique like Attribute-Oriented Induction (AOI), which drops keys identifying each tuple in the task-relevant data, produces generalised logic rules. An SQL-like operator for performing high-level queries on the original database and the rule base using key-preservation (KP) from AOI-KP is presented. Thus a combination of query processing on the database using rules and preserved keys from AOI-KP presents a useful and efficient mechanism for understanding the mined rules.