Choosing the right classifier for a RS is not easy and is in many senses task and
data-dependent. In the case of CF, some results seem to indicate that model-based
approaches using classifiers such as the SVM or Bayesian Networks can slightly
improve performance of the standard kNN classifier. However, those results are nonconclusive and hard to generalize. In the case of a content-based RS there is some
evidence that in some cases Bayesian Networks will perform better than simpler
methods such as decision trees. However, it is not clear that more complex nonlinear classifiers such as the ANN or SVMs can perform better.