Abstrato

A Roadmap to an Enhanced Graph Based Data mining Approach for Multi-Relational Data mining

D.Kavinya

Multi-relational data mining (MRDM) is a subfield of data mining which focuses on knowledge discovery from relational databases comprising multiple tables. The approaches to MRDM are categorized into five groups. They are greedy search based approach, inductive logic programming (ILP) based approach, inductive database based approach, mathematical graph theory based approach and kernel function based approach. Representation is a fundamental as well as a critical aspect in the process of discovery. These five approaches are grouped into two forms of representation, namely the graph-based representation (GBR) and the logic-based representation (LBR). In this paper, some major studies in each category are described, and representative methods among the studies are sketched. The comparison of Graph-based and logic-based multi-relational data mining according to the factors such as Structural Complexity, Semantic Complexity, Background Knowledge intended to condense the hypothesis space, and Background Knowledge intended to augment the hypothesis space is described and based on the comparison results the necessity of enhancing the Graph-based representation is enforced.