Method and system for detecting deviations in data tables

US6954756B2 · kind B2 · utility

Assignee

• International Business Machines Corporation · US

Inventors

• Andreas Arning · Rottenburg am Neckar, DE
• Toni Bollinger · Weil der Stadt, DE
• Reinhold Keuler · Waldenbuch, DE
• Friedemann Schwenkreis · Echterdingen, DE

Key dates

• Filing date: Aug 16, 2001
• Grant date: Oct 11, 2005
• Expiry date: Feb 19, 2023

Classification

• Technology area (CPC Y): Emerging Cross-Sectional Technologies
• CPC primary: Y10S707/99942
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

A method and system for automatically detecting deviations in a data table comprising a multitude of records and a multitude of columns. A column of the data table is selected as a classification column and a classification tree is calculated with respect to the classification column. Each edge of the classification tree is associated with a predicate. The leaf nodes of the classification tree are associated with a leaf record set comprising the subset of records for which the class predicate comprising all predicates along a path from a root node of the classification tree to the leaf nodes evaluates to TRUE. Leaf nodes are associated with a leaf label representing an expected value in the classification column for the corresponding leaf record sets. From the leaf record sets all records deviating with respect to the corresponding classification column from the leaf label are determined as deviation sets.