Analysing Factors That Influence Alumni Graduate Studies Attainment with Decision Trees

In Mexico, higher education is constantly suffering from low percentage of placement and interest of individuals for a graduate degree. Mexico needs more postgraduate students to increase the research and development activities and boost innovation in the private sector, especially in strategic industries. This paper suggests the use of data mining techniques to explore alumni factors and understand if these have a relationship with the alumnus returning to study a postgraduate degree. Fifteen attributes obtained from an alumni survey study were analyzed; this survey contains information from 12,780 former students, which graduated from a bachelor's degree in Tec de Monterrey. The Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology is used, and the machine learning algorithms, Random Forest, J48 and REPTree are compared to identify the best approach to build a classification model which can predict whether an alumni will study or not a postgraduate degree. For the purpose of this research, the data mining tool used was the Waikato Environment for Knowledge Analysis (WEKA). The resulting model shows that random forest outperforms the other decision tree algorithms based on the accuracy and classifier error, which drives the conclusion that this is a more suitable classifier for the explored dataset.