Data Mining Models for Student Databases

This thesis presents a data mining methodology to analyze the careers of University students, where a career is the ordered sequence of the exams taken by the single student. We present different models based on clustering, classification and sequential pattern techniques in order to understand and improve the performance of students and the scheduling of exams. We introduce an ideal career as the career of an ideal student which has taken each examination just after the end of the corresponding course, without delays. We then compare the career of a generic student with the ideal one by using these different techniques. Finally, we apply the methodology to a real case study and interpret the results which underline that the more students follow the order given by the ideal career the more they get good performance in terms of graduation time and final grade. We also analyze the career of university students from another point of view, that is, we can study the perspective of each course, by analyzing the distribution of students with respect to the delay with which they take an examination, to discover common characteristics between two or more courses. This is done in terms of Poisson distributions.