Improving Classification Knowledge Using an Integrated Knowledge Discovery Approach

Attribute-oriented induction approach (AOA) has been developed for knowledge discovery in large relational database. Several kinds of knowledge, such as characteristic rules and discrimination or classification rules can be discovered. These rules may contain unnecessary conditions and/or unnecessary conditionvalues. A Tuple-oriented approach (TOA) examines one tuple at a time since there are large number of possible combination in such testing, this approach is quite inefficient when performing learning from large databases. This paper introduces an Integrated Discovery System (IDS) based on combining techniques from both the AOA and the TOA. It captures the advantages and overcomes the difficulties associated with each of these approaches when used separately. Therefore IDS discovers more efficient classification rules, compared with the rules discovered by the pure AO.