La société Rapid-I, à travers leur logiciel phare RapidMiner, est un acteur très dynamique du l'informatique décisionnelle. Au-delà de l'outil, elle propose des solutions et des services dans le domaine de l'analyse prédictive, data mining et du text mining. Son site web regorge d'informations (blog, tutoriels, vidéos, forum, newsletter, wiki, etc.).
La version 5.0 de RapidMiner (Community Edition - Téléchargeable gratuitement) propose une interface profondément remaniée, s'inspirant visiblement de Knime. Les ressemblances entre les deux produits sont frappantes. Je me suis dit qu'il était opportun d'étudier cela en détail, evaluating its behavior in the context of a typical analysis. We hope to implement the following process: (1) construct and display a decision tree from a set of labeled observations, (2) save the tree in a file format PMML for later deployment and (3) assess the generalization performance of the classifier through cross-validation, (4) use the model to classify a set of unlabeled observations contained in a second file, the results (and label descriptors assigned) must be recorded in a third file in CSV format. These are
is very traditional tasks of data mining. We have repeatedly described in our courseware (eg SPAD ...). More reason to check it is easy to carry them out with this new version of RapidMiner. Indeed, with the previous version, some sequences were complicated. Establish a cross-validation, for example asking an organization, albeit a very rigorous in his mind, but not very intuitive.
Tags: RapidMiner, Knime, cross-validation, decision trees, deployment
Tutorial: fr_Tanagra_RapidMiner_5.pdf
Data : adult_rapidminer.zip
References:
Rapid-I, " RapidMiner "
0 comments:
Post a Comment