A data-mining algorithm discovers previously unknown patterns from datasets like graphs, web, and databases. Though there are various ways to store data, the most prominent and favorable one is database systems. Most of the important information is stored in the form of tables within the database system. Though information in a database is very robust and useful, it is not very interpretive. To extract knowledge from the database, we need to mine the data present in the database system.; The main objective of the thesis is to develop algorithms for converting any relational database (e.g., Oracle) into graph form for input into the Subdue graph-based data mining system. We evaluate such an algorithm both in terms of its scalability as the relational database increases in size and the effectiveness of the conversion as enabling Subdue to find patterns known to exist in the original database.
展开▼
