Implementing data mining in a DBMS

We have developed a clustering algorithm called CLIMIS to demonstrate the advantages of implementing a data mining algorithm in a database management system (DBMS). CLIMIS clusters data held in a DBMS, stores the resulting clusters in the DBMS and executes inside the DBMS. By tightly coupling CLIMIS with the database environment the algorithm scales better to large databases. This is achieved through an index-like structure that uses the database to overcome memory limitations. We further improve the performance of the algorithm by using a technique called adaptive clustering, which controls the size of the clusters.