Discovering Informative Patterns and Data Cleaning

Wepresent a methodfor discovering informative patterns from data. With this method,large databases can be reducedto only a few representative data entries. Ourframeworkencompassesalso methodsfor cleaning databases containing corrupted data. Bothon-line and off-line algorithms are proposedandexperimentally checkedon databases of handwrittenimages. Thegenerality of the framework makesit an attractive candidate for newapplications in knowledgediscovery. Keywords-knowledgediscovery, machine learning, informative patterns, data cleaning, informationgain.

• Type: Conference
• Year: 1994
• Where: KDD
• Authors: Isabelle Guyon, Nada Matic, Vladimir Vapnik
• Keywords: Data Mining, Databases, Informative Patterns, KDD 1994, Representative Data Entries