Automatic Classification of Attributes Using Feature Analysis
Original Publication Date: 2004-Aug-04
Included in the Prior Art Database: 2004-Aug-04
Database integration and migration are importatnt, but labor-intensive tasks. To tranform data from one representation to another, an expert user must identify and express correspondences between different attributes of different schemata. There is potentially an enormous number of attributes in a source schema that might correspond to a particular target attribute. The aim of this invention is to ease the burden on the user by classifying source attributes so that thet can be autoimatically and intelligently matched to target attributes. For categorical data, a novel variant of existing classfication techniques based on domain independent feature selection is presented. For numerical data, a quantile based classification method is presented, discovering characteristic distributions of data.