Abstract:

Scholar desertion is a main issue in superior educational institutions, analyzing the data of students for determining if certain patterns or behaviors can be found for determining possible cases of students which are or could be in risk for dropping their students could be very important for both, students and institutions.

At Mexican Universidad Autónoma Metropolitana, each scholar period (quarter), for different purposes, is prepared the Students General File (AGA by its spanish acronym) data set that permits the analysis of historical data of students. It contains personal information, grades of each registered subject, year of admission, accumulated credits, and a state, among several other data. The state of a student can be enrolled with registered subjects, enrolled without registered subjects, not enrolled, decommissioned for six quarters not enrolled, decommissioned for more than ten years as a student, and graduate.

Considering the importance of detecting students at risk of dropping their studies, this work presents an analysis of the trajectories of the Computer Engineering students with state of enrolled without registered subjects in the Azcapotzalco campus and presents an analysis of the nuanced course-taking patterns. 69 students and 3,592 records were analyzed, years of admission were from 2008 to 2019, covering the ten years that marks the institution's regulation.

Was analyzed the distribution of students by year of admission, term of admission, trajectories according to the credits accumulated per student, trajectories of students by year of admission, trajectories of students by gender, and comparison of the trajectories with that suggested by the curriculum, checking the dependency (aka, correlation) to see if autoregression models are applied.

This group of students pays for their registration, so they are entitled to most regular student services. For the program, these students are risky since, on the one hand, they decide not to study any subject, but on the other, they are interested in maintaining their commitment to the university. Therefore, it is essential to detect these possible cases to provide them with the necessary support to finish their studies.

Results show that several similarity metrics can be applied to identify students before they disaffection their studies. In such a way, apply many of the University support programs to those who need it most and obtain better efficiency indexes

Keywords:

Analysis of trajectories, risk students, autoregression models, similarity metrics, educational analytics.