Abstract:

Student dropout remains a critical challenge in the Higher Education Community. Recently, a comprehensive institutional study was conducted using a large sample of 3,099 Greek university students in order to validate a self-reported dropout risk measure, the Greek Dropout Tool – APrISE-15. Based on this initial dataset, the primary objective of the present work is to apply data mining techniques to the same institutional sample to construct and analyze distinct student profiles and to examine the statistical relationship between these profiles and the previously measured APrISE-15 risk scores.

The methodology focused on the structure of student profiles based on 22 mixed-type variables, which include demographics characteristics such as gender, age and year of study, digital activity (dimensions of online activity), and frequency of social interactions concerning extracurricular leisure time activities. The student profiling was conducted in two stages:
(1) Factor Analysis of Mixed Data (FAMD) for dimensionality reduction and handling of mixed data (quantitative and categorical variables), followed by
(2) k-means clustering on the resulting principal components to identify distinct student profiles, yielding four strong and distinct student profiles:
1. Consumer-Oriented Socialites,
2. Balanced Learners,
3. Socially Active Students, and
4.Socially Engaged Learners.

The subsequent analysis assessed the association between these four profiles and the APrISE-15 risk scores (total risk and the five sub-factors: Academic, Personal, Institutional, Economic, Social) applying Analysis of Variance (ANOVA) and post-hoc analysis. The results demonstrated that this student profile was a strong, predictive factor associated with variations in all risk dimensions. The Socially Active Students profile consistently presented the highest levels of self-reported risk in the academic and personal dimensions, while the Socially Engaged Learners profile displayed the lowest vulnerability. At the same time, the Consumer-Oriented Socialites exhibited a moderately elevated tendency toward academic disengagement, while the Balanced Learners reported relatively low risk levels, maintaining a well-regulated balance between academic, athletic and social dimensions. These findings strongly suggest that non-academic lifestyle profiles are significantly associated with self-reported dropout tendencies. This innovative profile model offers a valuable, context-specific tool for identifying high-risk groups, and facilitates the planning of timely targeted interventions in the institution under study.

Keywords:

Drop out, APrISE-15, profiles, university students.