One of the main problems with Adaptive Educational Hypermedia Systems (AEHS) is that is very difficult to test whether adaptation decisions are beneficial for all the students or some of them would benefit from a different adaptation. Data mining techniques can provide support to overcome, to a certain extent, this problem. This paper proposes the use of these techniques for detecting potential problems of adaptation in AEH systems. The proposed method searches for symptoms of these problems (called anomalies) through log analysis and tries to interpret the findings. Currently, a decision tree technique is being used for the task.