| Abstract | This technical report discusses the experimental comparison of commonly used algorithms both in their traditional discriminative form and as generative classifiers. The performance is compared using cost curves to see what benefits might be gained by using a generative classifier when the misclassification costs, and class frequencies, are unknown. There is some evidence that learning a discriminative classifier is more effective for a traditional classification task. Focusing on algorithms that have generative and discriminative forms, allows a clear comparison between these two types of classifier without being obscured by algorithmic differences. The report compares the performance of the classifiers over 16 data sets and for the full range of misclassification costs and class frequencies. The experiments show that there is some merit in using generative classifiers for cost sensitive learning but more work is needed to make them as effective as using multiple discriminative classifiers. |
|---|
