STAC is a Web platform that facilitates the application of statistical tests for the comparison of algorithms. Since the correct election of a test depends of a number of factors, such as the data distribution (parametrical or non-parametrical tests), the relationship among them (paired or unpaired) or the number of algorithms to be compared (pairwise or multiple), STAC includes the necessary help for deciding among the existing alternatives. Furthermore it has implemented an automatic assistant for suggesting which test should be used accordingly to the data available.
STAC includes the 13 most frequently used tests in this field. The platform allows to perform the comparison of algorithms of all scientific fields, such as artificial intelligence (machine learning, data mining, …) or other areas in the information technologies realm.
Authors
-
- Researchers
- Manuel Mucientes Molina
- Alberto Bugarín Diz
- Ismael Rodríguez Fernández
- Adrián Canosa Mouzo