Many students in the engineering disciplines do not complete their higher education degree and drop out. This problem is serious, especially for first-year university students. We analyse how students earn the ECTS credits required for their successful completion of the first study year.
We use data from the Faculty of Mechanical Engineering of the Czech Technical University, which offers the traditional classroom-based education. Given the university progression rules, we identify three groups of students: those who pass, those who earn just enough credits for staying in the program, and those who fail. Important patterns can be already found at the end of the first semester. We present an algorithm that identifies at-risk students. The purpose of this small project was to demonstrate that predictions followed by tutors' interventions do increase students' chances to progress in their study and improve the retention rate for the university.
Data of four consecutive years were used. In the academic year 2013/14, data of 994 first-year students were used to develop the predictive model. No interventions were provided and 33.2% students failed. The model was verified using 2014/15 data of 917 students - again without any intervention. The drop-out rate was 41.1%. In 2015/16 the model was for the first time applied to the cohort of 769 students. Predictions followed by three interventions by tutors were made at the end of the semester. The drop-out rate decreased to 16.9%. The same process was reapplied in the 2016/17 academic year, moreover a simple study planner/recommender has been developed to guide students.
Three interventions were provided early in the semester and an additional one before the end of the exam period. The drop-out rate was 16.71%.
The letter of recognition from the dean of the Faculty of Mechanical Engineering, Czech Technical University in Prague, informing about increase of the retention thanks to the predictions calculated by the system and well-targeted interventions.