The relational model has attracted wide-spread interests among database practitioners and researchers for its conceptual and for its mathematical elegance.
In this thesis, the study of dependencies between sets of attributes is performed for the utility of a Relational schema design.
Because of inadequency of functional dependencies in many-to-many relationships, a new concept of dependencies, that is, multivalued dependency is introduced and investigated its formal properties such as complete inference rules, BCNF, 4NF and so on. Related problems with multivalued dependencies (especially, membership problem, etc.) are also studied.
The problem related to decompose a relation into two projections without loss of information is considered.
Finally, efficient algorithms for the above problems are investigated and implemented.
본 논문은 Relational database 의 schema 를 design 할때 이용될 attribute 사이의 관계 즉 dependency에 관한 고찰로서, many-to-many 관계에 대한 Functional dependency 의 불충분을 해결하기 위해서 최근에 Fagin과 Zaniolo 가 소개한 Multiualued dependency (MVD) 에 대한 특성을 (Complete axiomatization, BCNF, 4NF 등) 연구하였다. Multivalued dependency 는 Functional dependency 를 확대하여 일반화 시킨 것이다.
특별히, 이것을 schema design 에 이용할때 생기는 membership 문제와 information의 loss 없이 Relation을 decompose하는 문제 (즉 nontrivial MVD 결정문제)를 연구 하였다. 이러한 문제들에 대한 efficient한 algorithm 및 implementation이 연구 되었다.