A New Method and Case Study for Predicting Tutoring Performance in Students at the Universidad Politécnica Salesiana using Data Science and Support Vector Machines

What has become evident over time in higher education is the low performance of students, especially in the first cycles, and higher education is of vital importance for our society today. That is why the GIETAES Group and ASU AyudantĀas Estudiantiles of the Universidad Politécnica Salesiana (UPS) offers tutoring to students. These tutorials are provided from students to students; however, the challenge is to detect the students most prone to fail the subject, in order to help them at an early stage. Therefore, in the present work, we propose an innovative analysis method of machine learning built in phases with the aim of predicting whether a student will lose or not the subject; firstly, we perform data preparation in which a preprocessing of variables, variable analysis, secondly, we perform a predictive analysis for this we have experimented with some techniques including support vector machines, Random Forest (RF) algorithm, KNN algorithm, and finally in the third phase performs the evaluation and interpretation of results. To demonstrate the effectiveness of our method, we have used a UPS own dataset and evaluated it with several quality metrics such as accuracy, precision, recall, and F1-Score. This research is a base point to experiment with various parameters based on the low performance of students not only in higher education but in any educational entity.