H.: Stand on the Shoulders of Giants: Towards a Portal for Collaborative Experimentation (2009)
| Venue: | in Data Mining. International Workshop on Third Generation Data Mining at ECML PKDD |
| Citations: | 2 - 0 self |
BibTeX
@INPROCEEDINGS{Vanschoren09h.:stand,
author = {Joaquin Vanschoren and Hendrik Blockeel},
title = {H.: Stand on the Shoulders of Giants: Towards a Portal for Collaborative Experimentation},
booktitle = {in Data Mining. International Workshop on Third Generation Data Mining at ECML PKDD},
year = {2009},
pages = {88--89}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract. Data mining is a science guided by experimentation. Still, there exists only limited support to automate such experimentation, making it a laborious and error-prone process. In this paper, we plead for a more collaborative approach in which researchers build upon prior efforts by other researchers in setting up and running experiments, and in turn share the fruits of their experimentation with the community. To this aim, we propose an online portal: a web service which collects anno-tated datasets, algorithms, experimental setups and millions of empirical evaluations sent to it by data mining tools and individual users, organizes them in one searchable central resource, and assists researchers in set-ting up and running new experiments. Such a service could drive many useful applications, from online platforms where all stored experiments can be explored in great depth and discovered trends can be discussed, to advanced data mining tools that interact with the service to speed up the setup and execution of experiments and upload new experiments to allow a more in-depth exploration of the acquired results. 1
Keyphrases
collaborative experimentation data mining tool in-depth exploration prior effort individual user new experiment web service searchable central resource experimental setup collaborative approach online portal anno-tated datasets great depth upload new experiment online platform data mining limited support turn share error-prone process many useful application empirical evaluation
