1. Matrices created from test’s data. We have applied the Apriori-CAR mining algorithm over the previously described data matrices. In the first experiment, we have used the score-matrix, and we have selected/filtered as input-attributes (antecedent) only the item answers, and as class the final score. In this way, we can see the relationships between items and how they can predict/determine the final score obtained by students. In the second experiment, we have used the knowledge-matrix, and we have selected the knowledge of concepts as input- attributes (antecedent) and the final score as a class attribute. In this way, we can discover the relationships between concepts and between the level of knowledge of these concepts and the final score obtained by students. 3 Conclusions and Future Work. In this paper, we proposed to use a special type of association rules over the assessment data in a particular scenario. We mined different test data matrices rather than only the typical score-matrix. Particularly we used an item-concept relationship matrix created by the instructor and a student-concept knowledge level matrix automatically created based on the information from the other two matrices. Finally, it is important to notice that concepts themselves may need to be presented as a hierarchy rather than a 'flat' set of independent concepts. Mining interesting patterns in such settings is one of the directions of our further work.