MODELING THE PREDICTION OF STUDENTS' SUCCESS IN THE CONTEXT OF SMALL UNIVERSITIES

The prediction of the success of college students becomes one of the most important but at the same time very demanding themes of university research. Early detection of students at risk of failure is of great importance both for students themselves and for universities that seek to reduce students' failure in courses leading to their early school leaving.

Researchers in this field involve methods from the field of classification and regression algorithms or probability models. Frequent interest of researchers is the orientation in the large amount of data currently available to universities through datamining methods.

The aim of this paper is to find a method for predict the success of students in the environment of small universities. In this environment, we encounter mainly the problem of the low number of students associated with a small range of the group. This is a case where commonly used methods fail and it is necessary to look for specific approaches that would allow predictions on such limited data. On the other hand, a small number of students give these universities a great advantage in the form of a very effective intervention. Finding suitable methods for modeling student success is therefore very beneficial in this environment.