Industry applications of data mining: challenges and opportunities

Abstract

Summary form only given, as follows. Data mining applications deployed in industry are aimed at satisfying two problems organizations face: customer intimacy and better utilization of data assets. These applications can be divided into those that use micro-mining, i.e. single-mining-component desktop systems, and those who use macro-mining, i.e. multi-component server-based systems. The macro-mining applications are usually coupled with data warehouses. The interesting result of this coupling for the data mining community is that the data warehouses cannot be supported by the current data mining offerings delaying the deployment of applications in production environments. The data volumes are too large, the data types too diverse and the data characteristics too incompatible for the existing data mining algorithms. Furthermore, the pure mining operation is a very small part of the entire application life-cycle. The author presents the issues related to the coupling of macro-mining with data warehouses, and proposes issues that must be resolved for large-scale data mining applications to continue being deployed successfully.