Metaclassify

If you don't know what is Weka, the present tool is maybe not for you; but if you generate huge quantities of biological data you should consider looking at it. Weka is a very powerful toolbox giving access to several state-of-the-art data mining algorithms that you can apply on your dataset. The main goal is to identify patterns (rules, decision trees, regression models) in your data that link attributes between them, or link attributes to a class.

One of the problem usually encountered with Weka (if you're not a machine learning guru) is to choose the 'best' classifier, the one able to find patterns with the higher accuracy, support, or any other score you have in mind to judge it. MetaClassify has been made to help you to solve this problem. By providing a dataset in the .arff format, MetaClassify will

determine it's type (nominal or numeric class, nominal and/or numeric attributes; for now, string and date-type attributes are not supported)
automatically select the classifiers able to deal with this kind of dataset
launch these classifiers, ask for a 10-fold cross-validation, and retrieve the results
generate a report with all the scores (error on training data, and - for classification models - detailed accuracy per class and confusion matrix), plus an indication of the time took by each algorithm to generate a model

References:

Ian H. Witten and Eibe Frank (2005). Data Mining: Practical machine learning tools and techniques, 2nd Edition, Morgan Kaufmann, San Francisco
(note: I really, really urge you to read this book if you're interested by the domain of data mining)

Version

1.3 (Dec 21, 2006)

Documentation

The documentation is included in the package below, in format.

Download

Metaclassify 1.3 — December 21, 2006 Download (20.3 KB)