Software

I am pleased to announce the release of TunedIT system (http://tunedit.org/), which facilitates experimental investigation and scientific collaboration in the field of data mining.

TunedIT is an integrated system for sharing, evaluation and comparison of data-mining and machine-learning algorithms. Its aim is to help researchers and practitioners navigate in the wealth of algorithms, datasets and experimental results, discover true performance of different methods, plan and carry out new insightful research. Its website contains performance data for nearly 100 algorithms tested on numerous datasets - these include the algorithms from Weka and Rseslib libraries and the datasets from UCI Machine Learning Repository. Everyone is welcome to contribute new resources and results.

TunedIT is composed of three complementary modules: TunedTester, Repository and Knowledge Base.

TunedTester is a Java application for automated evaluation of algorithms and generation of reproducible results. It runs locally on user's computer. All resources necessary to set up a test are automatically downloaded from Repository. If requested, TunedTester can submit results to Knowledge Base. Thanks to automation of tests, results generated by different users are all valid and can be easily reproduced by anyone else.

Repository is a public collection of data-mining resources, open for everyone. The resources include, among other things, the algorithms, datasets and evaluation procedures which can be used by TunedTester to set up and execute experiments.

Knowledge Base is a database of test results generated by TunedTester. It merges all the results submitted by different users into rich and comprehensive database of performance profiles of different algorithms evaluated on a wide range of datasets. Users may browse its contents with convenient web interface and download the results as CSV files for off-line analysis.

TunedIT components contain numerous security measures to ensure that all results stored in Knowledge Base are valid, no matter who submitted them. Knowledge Base cannot be polluted neither by accidental mistakes nor intentional fakery of any user.

TunedIT may help researchers design repeatable experiments and generate reproducible results. It may be particularly useful when conducting experiments intended for publication, as reproducibility of experimental results is the essential factor that determines research value of the paper. TunedIT helps also in dissemination of new ideas and findings. Every researcher may upload his implementations, datasets and documents into Repository, so that other users can find them easily and employ in their own research.

I invite you to visit and explore TunedIT, as well as contribute new contents. TunedIT website: http://tunedit.org/

Marcin Wojnarski Project Lead TunedIT