Today directed random simulation is one of the most commonly used verification techniques. Because this technique in no proof of correctness, it is important to test the design as complete as possible. But this is a hard to reach goal, that needs a lot of computing power and much human interaction. There has been a proposal for using Bayesian networks to implement an automatic feedback loop (Shai Fine et al, 40th Design Automation Conference, 2003). In addition, this paper introduces another implementation of an automatic feedback loop using data mining techniques. Both approaches are applied to the same design and the results are compared.