An Information-Theoretic Approach to Data Mining (1999)
| Citations: | 3 - 3 self |
BibTeX
@MISC{Last99aninformation-theoretic,
author = {Mark Last and Oded Maimon},
title = {An Information-Theoretic Approach to Data Mining},
year = {1999}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
1. Introduction The Knowledge Discovery in Databases (KDD) process is defined by Fayyad et al. [5] as "the nontrivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data". Since the potential number of patterns presenting in a real-world database (containing, at least, tens of attributes and thousands of records) may be nearly infinite, most data mining techniques are concerned with a computationally-efficient enumeration of patterns (models, causation relations, association rules, etc.) to discover the most useful patterns for the user or the task. Our work aims at finding a minimal set of database attributes involved in discovered patterns by using an information-theoretic statistical approach. The problem of minimizing the number of attributes is important for several reasons. First, data mining requires a significant effort associated with data preparation (including the rapidly evolving approach of data warehousing), and this e...
Keyphrases
data mining information-theoretic approach data mining technique nontrivial process minimal set significant effort real-world database potential number several reason knowledge discovery data preparation causation relation work aim information-theoretic statistical approach understandable pattern association rule computationally-efficient enumeration useful pattern database attribute
