Example Queries

Some examples of how to use the database

Click to retrieve the query, copy-paste it into the query field to run.
See all stored queries

Comparing learners

• Comparing all algorithms in the database on a specific dataset D
• Comparing all algorithms in the database, on a specific dataset D, and distinguish between baselearners used in ensembles and kernels used in kernel methods
• Directly compare two algorithms on all datasets

Ranking learners

• Order all algorithms by the average of their maximum performance (over different parameter settings) on each dataset
• Compare algorithms over all UCI datasets, using a range of evaluation metrics, all normalized between baseline and maximum performance
• Ranking of algorithms over all UCI datasets, using the Friedman rank and scaled by the critical difference as calculated by the Nemenyi test
• Ranking of algorithms, including base-learners and kernels, over all UCI datasets, using the Friedman rank and scaled by the critical difference as calculated by the Nemenyi test

Algorithm parameters

• Studying the effect of a parameter P on algorithm A on dataset D
• Compare the effect of a parameter P from algorithm A on a range of datasets

Dataset properties and preprocessing

• Show the combined effects of a parameter P and dataset property DP on algorithm A over all UCI datasets
• Draw the learning curve for algorithm A on dataset D
• Show the effect of data property DP on the optimal value of parameter P
• Show the value of a data property DP, ordering datasets by time of publication
• Show the performance difference of two algorithms on datasets ordered by date of publication

Algorithm properties

• Calculate bias-variance profile of each algorithm using a variety of B/V measures
• Calculate bias-variance profile of each algorithm using Kohavi-Wolpert's definition, and differentiate between different baselearners in ensembles and kernels in kernel methods
• Show the impact of data property DP on the percentage of bias error in the total error for a number of different algorithms