Machine Learning Model Applied to Higher Education

Abstract

Student desertion is one of the main social problems around the world. Consequently, to propose this issue, there are several studies under different circumstances or scenarios. For this reason, this research creates four datasets, which take the common variables of easy extraction to the academic process. These variables have been grouped under common characteristics such as general student profile information, admission process information, financial information, academic information, and academic performance information. Thus, the method used in this research is analytical, since it is intended to analyze each subset of data in order to identify the variables with the greatest impact on university dropout. As a result, have been identified the variables with impact on university dropout, For this, a neural network model has been implemented using Python and Keras. In conclusion, the research evidence that academic information is mostly related to college dropout related to university dropout, while admission, financial, and student profile information are not significant in detecting or predicting college dropout. However, with the data obtained, it has been shown that the prediction is not early but in many cases late, since the notes would have already been delivered to students. Therefore, future research is intended to identify the causes that originate academic problems.