engleski

Enhancing the Efficiency of Data Mining Process by Circular Coupling of Multiple Schemes

A widespread proliferation of databases has created a demand for powerful tools for turning stored data into useful knowledge. The paper explicates a data classification and description environment that combines data mining procedures in a novel style, not encountered in standard architectures. The environment (implemented in Java) is primarily designed for classification and description problems that exhibit a high degree of nondeterministic behavior. The architecture utilizes SQL queries for direct database access. In the preprocessing phase, extracted data is analyzed using greedy hill-climbing based algorithm that helps in selecting the most promising attribute set. For the classification purposes a Naive Bayes classifier is used, while for the dataset description purposes a C4.5 type decision tree generator is implemented. An embedded two-dimensional visual performance evaluation tool enables rapid evaluation of the extracted knowledge. The predictive power of the described data mining environment is illustrated on a typical real-world business application.

Automated reasoning; Data mining; Knowledge discovery; Probabilistic reasoning

nije evidentirano