A direct measure of discriminant and characteristic capability for classifier building and assessment

摘要: Performance measures are used in various stages of the process aimed at solving a classification problem. Unfortunately, most these fact biased, meaning that they strictly depend on class ratio - i.e. imbalance between negative and positive samples. After pointing to source bias for best known measures, novel unbiased defined which able capture concepts discriminant characteristic capability. The combined use can give important information researchers involved machine learning or pattern recognition tasks, particular classifier performance assessment feature selection.