In recent years, two closely related research communities—educational data mining (EDM) and learning analytics (LA)—have emerged around a shared interest in leveraging large-scale educational data to enhance teaching, learning, and institutional decision making. These communities draw on advances in data science and machine learning to develop models that can describe, explain, and predict student behaviour and performance within educational systems. Among the wide range of analytical techniques employed in this domain, decision tree classifiers have gained prominence due to their interpretability, ease of implementation, and ability to uncover meaningful patterns in complex datasets. This study contributes to the growing body of work in EDM by examining the use of a decision tree algorithm to model and predict student academic outcomes in a South African higher education context. Furthermore, this study directly informs the teaching and learning and modes of delivery at institutions of higher learning with similar contexts.

The study draws on two primary sources of data. The first dataset consists of a cohort of 2,204 students enrolled in Science, Technology and Engineering programmes at a South African university who completed three higher education admission tests of academic literacies at the start of their undergraduate studies. These assessments are widely used in the South African context to evaluate students’ academic preparedness and to inform admission (selection and placement) and curriculum decisions. The second dataset captures student performance in higher education through a longitudinal measure of academic standing. Academic standing reflects a student’s academic progression over time and includes outcomes such as continuation, graduation, or exclusion. For the purposes of this study, academic standing is operationalised into two broad outcome categories: has a student Graduated (GRAD) or not (NOTGRAD). The final academic standing used in the analysis corresponds to each student’s status at the end of regulation time for graduation, regulation time plus one year, and regulation time plus an additional two years, thereby allowing for delayed completion while still distinguishing successful from unsuccessful outcomes.

The primary objective of the study is to investigate whether performance on higher education admission tests of academic literacies can be used to build a predictive model of long-term academic outcomes using a decision tree algorithm. To this end, the full cohort is divided into sub-cohorts, with one subset used to train the model and another reserved for validation and prediction. Predictor variables include selected measures of prior academic performance derived from three admission test results in Academic Literacy, Quantitative Literacy and Mathematics. The decision tree algorithm is applied to the training data to identify hierarchical decision rules that distinguish between students who are able to graduate or not (that is either drop out, or remain enrolled) within the defined timeframe.

The results demonstrate that the decision tree model can classify students into the two academic standing categories with a meaningful level of accuracy. In particular, the model shows strong potential in identifying students who are at risk of not graduating, as well as those who are likely to graduate within the regulation time, regulation time plus one year, and regulation time plus two years. The transparent structure of the decision tree allows for clear interpretation of the relationships between prior academic performance indicators and later academic outcomes, making the model especially suitable for use by institutional stakeholders such as academic planners, advisors, and teaching staff.

From a practical perspective, the findings highlight the value of using pre-entry academic literacies assessment data to inform proactive institutional teaching and learning interventions. Accurate prediction of student performance enables higher education institutions to respond timeously to the diverse needs of students through targeted placement decisions, academic support programmes, and curricular interventions. In contexts such as South Africa, where higher education systems face persistent challenges related to access, equity, and student success, predictive models of this nature can play an important role in improving retention and graduation rates.

While the study confirms the feasibility of decision tree algorithms for predicting student outcomes, it also highlights areas for improvement. Future research could improve predictive accuracy by incorporating real-time engagement data (e.g., LMS activity and attendance) and by using larger, more diverse cohorts. Overall, the findings underscore the importance of interpretable, data-informed approaches to supporting student success in higher education.

In the past 25 years, the number of students enrolled at private higher education institutions has risen from 1% to 13% (Destatis 2025). Who are the students enrolling in private higher education institutions, and what do they expect from their institutions? Students could study a comparable program at a public higher education institution and not pay any tuition fees. Why do they pay tuition fees and what do they expect in return? Our research question is:

What do students expect from studying at a private Higher Education Institution in Germany?

We answer the research question using interview data and a survey of students at private higher education institutions in Germany, which we collected from June 2024 to March 2025. The study is exploratory in nature and therefore does not seek to test theoretically generated hypotheses. In our analysis, we focus on students studying at private universities of applied sciences (UAS).

The most important finding we made was the difference in expectations between students enrolled in on-campus programs and those enrolled in distance learning programs.

The German private higher education system and its students

In Germany, a distinction is made between private research universities and private universities of applied sciences. We base our distinction on the classification used by Statistisches Bundesamt (Destatis) (2025) for the collection of higher education statistics. Using this data, we show that private universities of applied sciences appeal mostly to a very specific group of students: working students who are older, want to study alongside their work, and also have children to care for.

Methods

The data evaluated comes from the BMFTR project “ELLpH – Influencing factors on new teaching and learning concepts at private higher education institutions.” In this project, seven private higher education institutions were selected according to the typology of Frank et al. (2020) in order to cover a broad spectrum. The data is based on an interview study with 18 students and 14 alumni. In addition, a quantitative survey of students at six private higher education institutions was conducted between January and March 2025. Five of the universities are UAS and one is a research university. Since the study programs offered and the students themselves differ greatly between research universities and UAS, we have limited the following analyses to UAS and their students. The entire data set encompasses 508 fully completed questionnaires, with 449 from UAS.

Empirical evidence: expectations from students

In both the qualitative and quantitative studies, we found a significant difference in students' expectations depending on the type of study program: whether it was an on-campus or distance learning program. For this reason, we compare the expectations of students who come from distance learning with those who study an on-campus program.

In addition to interesting insights from the qualitative interviews, we also describe the results of the quantitative survey. The answer to the following question serves as an example: “Why did you decide to study at the higher education institution where you are currently enrolled? Please tick a maximum of 5 of the most important reasons”. In distance learning flexibility is the top priority. The range of courses, learning times, and exam times should be as flexible as possible. In on-campus programmes students expect a highly structured, predetermined programme that closely integrates theory and practice. The programme should be organised in small groups and offer opportunities to build a career network. We also analysed the motivational differences and basic needs between the two groups. Relatedness is particularly important for on-campus students while autonomy is important for distance learners.

Distance learning is particularly popular among students from non-academic backgrounds and women. Many of them also have vocational training. However, the high workload experienced by those working more than 25 hours per week has little influence. These issues are likely to be adequately addressed in on-campus studies. The question here is whether this group is doubly disadvantaged: not only do they find it particularly difficult to succeed in the higher education system, but they also have to pay tuition fees.

Introduction

Student satisfaction has emerged as a dominant proxy indicator for evaluating student experience in South African higher education institutions in recent years. Although this metric provides utility in assessing service delivery, it remains insufficient for capturing the nuanced complexities of the lived experience within resource-constrained contexts. This paper proposes a conceptual and methodological shift from satisfaction-focused outcomes to a process-oriented assessment framework grounded in student-centredness. The objective is to cultivate a deeper connection between higher education institutions and their diverse student populations, thereby fostering institutional environments that enable student-centred approaches to teaching and support rather than merely measuring reactive satisfaction with services rendered.

This study is situated at a South African university which currently enrolls approximately 55,000 undergraduate(76%) and postgraduate(24%) students. The research is anchored in longitudinal data from the institution's Undergraduate Experience Survey (UGES), administered annually to capture systematic patterns in student experience. This study utilizes survey data spanning from 2022 to 2025, representing four consecutive iterations that permit longitudinal measurement across multiple years.

Theoretical Framework

To transcend the inherent limitations of satisfaction-based metrics, this study adopts the Capabilities Approach (CA), articulated by Sen (1999) and operationalized within the higher education context through Wilson-Strydom's (2015) university-specific capabilities framework. Whereas conventional satisfaction surveys measure students' subjective reactions to institutional services through utility-based metrics that assume consumer preferences, the Capabilities Approach instead assesses the substantive opportunities and freedoms (capabilities) that students possess to achieve valued functionings associated with becoming an educated person, thereby constituting a well-being-oriented metric grounded in human development principles (Nussbaum, 2011; Wilson-Strydom, 2015). Within the profoundly unequal landscape characteristic of South African higher education, satisfaction measures prove particularly unreliable due to the phenomenon of adaptive preferences, whereby students originating from historically marginalized and economically disadvantaged backgrounds may report satisfaction despite experiencing objective deprivation, primarily because structural inequalities have conditioned them to maintain diminished expectations regarding institutional support and educational quality (Sen, 1992).

Consequently, this study challenges the dominant neoliberal conception positioning the university primarily as a service provider responding to consumer preferences, advocating instead for a reconceptualization of higher education institutions as enabling environments that establish specific pedagogical, psychosocial, and material conditions allowing students to convert available resources and opportunities into successful academic and personal outcomes (Walker, 2020; Wilson-Strydom, 2017).

Proposed Latent Constructs (2022–2025)

To advance beyond univariate satisfaction metrics toward a multidimensional assessment of student-centredness, this study proposes five latent constructs derived from survey items that demonstrate consistency across the four-year longitudinal dataset. Pedagogical Care captures the relational quality of teaching interactions, extending beyond mere content delivery to encompass the human connection dimension of pedagogy, aggregating survey items assessing whether lecturers are "helpful," "approachable," "encourage discussions," and whether "at least one lecturer cares about me as a person." Assessment for Learning evaluates the developmental value of assessment practices through items addressing the clarity of assessment requirements, perceived fairness in marking procedures, and the timeliness and usefulness of feedback provided (Ross et al., 2024). Digital Interface Efficacy serves as both a substantive construct and a methodological control variable, measuring students' user experience of the Learning Management System across the institutional transition from Blackboard (2022–2023) to Moodle (2024–2025). Institutional Inclusion and Belonging directly addresses social justice imperatives by measuring students' psychosocial integration through indicators of physical and psychological safety, sense of belonging, and perceptions of institutional tolerance regarding diversity. Basic Needs Security constitutes a critical addition particularly salient within the South African context, where food insecurity affects between 30% and 38% of university students (Adeniyi & Durojaye, 2020), combining items assessing food security, financial anxiety, and satisfaction with critical support divisions.

Methodology

This study employs a quantitative longitudinal design utilizing secondary data from four consecutive administrations of the UGES spanning 2022 through 2025. Confirmatory Factor Analysis will be conducted using R statistical software (lavaan package) to test the validity of the five proposed constructs, evaluating model fit against established indices targeting a Comparative Fit Index (CFI) exceeding 0.95 and a Root Mean Square Error of Approximation (RMSEA) below 0.06 (Hair et al., 2019).

This study contributes to advancing both theoretical understanding and practical assessment of student experience by demonstrating the advantages of transitioning from unidimensional satisfaction metrics to a holistic, capabilities-informed student-centred assessment framework (Wahl et al., 2023).