Visual-Predictive Data Analysis Approach for the Academic Performance of Students from a Peruvian University

Abstract

The academic success of university students is a problem that depends in a multi-factorial way on the aspects related to the student and the career itself. A problem with this level of complexity needs to be faced with integral approaches, which involves the complement of numerical quantitative analysis with other types of analysis. This study uses a novel visual-predictive data analysis approach to obtain relevant information regarding the academic performance of students from a Peruvian university. This approach joins together domain understanding and data-visualization analysis, with the construction of machine learning models in order to provide a visual-predictive model of the students’ academic success. Specifically, a trained XGBoost Machine Learning model achieved a performance of up to 91.5% Accuracy. The results obtained alongside a visual data analysis allow us to identify the relevant variables associated with the students’ academic performances. In this study, this novel approach was found to be a valuable tool for developing and targeting policies to support students with lower academic performance or to stimulate advanced students. Moreover, we were able to give some insight into the academic situation of the different careers of the university.