An Attribute-Oriented Induction Approach for Knowledge Discovery from Relational Databases

摘要

Attribute-Oriented Induction (AOI) is one of the most important algorithms for Data Mining, which contains a relational database and a concept hierarchy (concept tree) for each attribute, and its outputs are small relations that summarize the generalized characteristics of the task-relevant data. However the concept trees will vary in different classification standards, not to mention AOI can only deal with data types of single-valued attributes, therefore the requirement arises immediately- that is how to design a new classification algorithm to classify the multi-valued attributes. In addition, it would be time-consuming in processing data mining and data warehousing manually. In view of these weaknesses, this paper proposes a Modified AOI (MAOI) algorithm that converts the data to Boolean bit, uses K-map to converge the attributes, and finally, enables us to discover the generalized knowledge from Relational Databases.