Paper Info
Title
Qualitative Assessment of Data-Mining Workflows.
Abstract
Data-mining is aimed at discovering knowledge from data. Advanced data-mining software, such as Weka, Orange and Rapid Miner, provide hundreds of data-mining methods. In order to perform a given data-mining task, these methods have to be selected and combined into a data-mining workflow. Traditionally, workflows are designed by data-mining experts, but this is difficult and there is a strong need to automate workflow design. In doing so, it is essential to be able to assess the quality of workflows. So far, this was usually assessed only through performance indicators, such as classification accuracy. In this paper, we present a workflow assessment model that uses an extended set of user-oriented indicators, which include understandability for the user, generality of used components, and robustness of the workflow. The model, which was developed using software DEXi, is qualitative, multi-attribute, hierarchical, and rule-based. We describe its components, current implementation of the model, and illustrate its performance on the case of two workflows.
Year
DOI
Venue
2012
10.3233/978-1-61499-073-4-75
Frontiers in Artificial Intelligence and Applications
Keywords
Field
DocType
data-mining workflows,qualitative assessment,multi-criteria modeling,understandability,robustness
Data mining,Computer science,Knowledge management,Workflow
Conference
Volume
ISSN
Citations 
238
0922-6389
1
PageRank 
References 
Authors
0.43
0
3
Name
Order
Citations
PageRank
Martin Znidarsic113311.51
Marko Bohanec233448.69
Nejc Trdin3102.45