What is ClowdFlows?
It is an online data science service: it enables the user to design and run data analysis workflows on a web browser, without installing any software. It provides lots of widgets from known libraries, e.g. Orange, SciKit, Weka. It can also call external web services. Processing is performed on our servers and can access data remotely from files or databases.
WorkFlows can be made public in order to include the URL in a publication to enable people to reproduce the experiments, and more generally to be used as tutorials. Examples are provided below.
It is an open-source software. It can easily be extended. Our contributions are listed below.
A user guide is available here.
Main differences between the nodes
- ClowdFlows.org is hosted in Ljubljana, Slovenia. ClowdFlows Unistra is hosted in Strasbourg, France.
- Clowdflows Unistra has a separated authentification and database.
- Improvements of ClowdFlows and new algorithms implemented in Ljubljana are hosted first on ClowdFlows.org. New algorithms implemented in Strasbourg are hosted first on ClowdFlows Unistra.
In the relational data mining package
- The rule discovery tool, Tertius. See the example: Tertius from files on the family example
- The first-order Bayesian classifiers 1BC and 1BC2. See the example: 1BC on mutagenesis using cross-validation from DBMS
- Propositionalisation using Relaggs, cardinalisation and quantiles. See the example: Compare propositionalisation and J48 on urban blocks