A Multi-Tier Architecture for High-Performance Data Mining

Abstract

Data mining has been recognised as an essential element of decision support, which has increasingly become a focus of the database industry. Like all computationally expensive data analysis applications, for example Online Analytical Processing (OLAP), performance is a key factor for usefulness and acceptance in business. In the course of the CRITIKAL1 project (Client-Server Rule Induction Technology for Industrial Knowledge Acquisition from Large Databases), which is funded by the European Commission, several kinds of architectures for data mining were evaluated with a strong focus on high performance. Specifically, the data mining techniques association rule discovery and decision tree induction were implemented into a prototype. We present the architecture developed by the CRITIKAL consortium and compare it to alternative architectures.