A hierarchical model learning approach for refining and managing concept clusters discovered from databases

摘要

The contents of most databases are ever-changing, and erroneous data can be a significant problem in real-world databases. Therefore, the process of discovering knowledge from databases is a process based on incipient hypothesis generation/evaluation and refinement/management. Although many systems for knowledge discovery in databases have been proposed, most systems have not addressed the capabilities of refining/managing the discovered knowledge. This paper describes a hierarchical model learning approach for refining/managing concept clusters discovered from databases. This approach is the basic one for developing HML (Hierarchical Model Learning), which is one sub-system of our GLS (Global Learning Scheme) discovery system and can be cooperatively used with other sub-systems of GLS such as DBI (Decomposition Based Induction). By means of HML, concept clusters discovered from a database by DBI can be represented as the Multi-Layer Logic formulae with hierarchical models in a knowledge-base and can be easily refined/managed according to data change in a database and/or domain knowledge. HML is based on the model representation of Multi-Layer Logic (MLL). Its key feature is the quantitative evaluation for selecting the best representation of the MLL formulae by using cooperatively a criterion based on information theory and domain knowledge. Experience with a prototype of HML implemented by the knowledge-based system KAUS is discussed.