Here's some sources of data which could be used with learning or data mining algorithms. If you want to do real statistically valid %comparisons, you need to be aware of potential problems. For a good introduction to these, see the paper "On Comparing Classifiers:Pitfalls and to Avoid and a Recommended Approach" by Steven Salzberg

• MLnet Machine Learning Archive Datasets
• The UCIRepository Of Machine Learning Databases And Domain Theories
• Stat Lib Datasets Archive
• University Of Toronto SDelve Datasets
• cell phone spy
• Statlog Datasets
• Kdd Sisyphus
• Synthetic Classification Data Sets
• Kaggle See also: Evaluation of Intelligent Systems - http://eksl-www.cs.umass.edu/eis/