Integração mineração de dados – SGBD não é uma panacéia: estudo da integração do algoritmo Apriori Quantitativo ao Oracle9i.

Abstract

Most early data mining algorithms were developed largely on specialized file systems, without the DBMS management. However, modern researches begun appeared fitting to couple the database systems and data mining techniques and summing the mining approaches up DBMS’s known advantages as transactions control and atomicity, security, robustness and query optimization. This work analyzes the Quantitative Apriori [4] algorithm, state of the art in association rules generation completely integrated to Oracle 9i Object-Relational DBMS. The algorithm almost never there is over the specialized literature, so it wasn’t be tested and used. The results (execution time) of it were compared to Apriori [2] classical association rules generation algorithm, either coupled to Oracle 9i,looking at analyses how first one would be higher than the last one. The results, although disturbing, let us conclude that couple the data mining algorithms to DBMS’s isn’t an enough solution. The problems that we had with the Apriori Quantitative integration to Oracle9i are discussed in details and the general tests conclusions are being deduced. The databases used for algorithms were modeled by two distinct forms: first, purely relational; second, with object-relational extensions, wearing, for example, collection types. Then, an other work purpose was analyze the data structures relational and object-relational influence about the algorithms performance, looking for general and useful conclusions about superiority of one then other.