Application Of Inductive Logic Programming To Knowledge Discovery In Databases(KDD)

My graduate thesis is Application of Inductive Logic Programming (ILP) to Knowledge Discovery in Databases (KDD), a part of a 863 Project. This paper is a summary of my work in the past two years. In the world data has increased along with rapid development of database technology and application of DBMS, So it becomes a challenging task to find out valuable knowledge in databases. The current DBMS can deal with task of transaction efficiently, but as to analysis process, that is finding relations and rules in databases, they lack efficient way to discover the knowledge hiding in databases and to foresee the future developing trend. KDD is a new research field, which find knowledge hiding in a large amount of data by coupling kinds of algorithms (most are Al algorithms) with DBMS. This thesis analyses three methods coupling inductive logic programming with the database, gives the relationships between database and predicates. And in the thesis we describe two prototype systems: the first is a loosing-coupling interface between DBMS and inductive logic programming, the second is a prototype KDD system (KDDGOL), on the clause level. Up to now, We have not seen other work on application of ILP to KDD in China. RDT/DB is the only work we know in this direction in the world. KDDGOL is a fresh try that couple ILP with DBMS. In the implementation we make some restrictions on mode declaration (Bias) considering efficiency and feasibility of the system, By doing so we weaken the power of the system. we also should enhance this system in future to deal with noise in databases.