Data Mining for Effective Risk Analysis in a Bank Intelligence Scenario

We propose a data warehousing architecture for effective risk analysis in a banking scenario. The core of the architecture consists in two data mining tools for improving the quality of consolidated data during the acquisition process. Specifically, we deal with schema reconciliation, i.e. segmentation of a string sequence according to fixed attribute schema. To this purpose we present the RecBoost methodology which pursuits effective reconciliation via multiple, stages of classification. In addition, we propose a hash-based technique for data reconciliation, i.e. the recognition of apparently different records that, as a matter of fact, refer to the same real-world entity.