Educational data mining for decision-making: a framework based on student development theory

Xiaofeng Lei, Ming Yang, Yi Cai2017 · DOI

Abstract

Educational Data Mining (EDM) is an increasingly vital field for university administrators seeking to optimize student outcomes. However, many current EDM approaches focus on narrow academic performance metrics without a grounding in broader student development theories. This paper proposes a comprehensive framework that integrates Alexander Astin’s "Inputs-Environment-Outcomes" (I-E-O) model and Student Involvement Theory into an iterative EDM cycle. By applying this framework to a dataset of 2,266 master’s students, the study demonstrates how university leaders can move from simple data collection to informed, theory-driven policy making.

Introduction

Higher education institutions are currently generating vast quantities of data, ranging from admissions records to post-graduation employment statistics. While this data holds the potential to transform institutional decision-making, it often remains siloed or underutilized. A significant challenge is the "Black Box" of student development: while administrators can see what students bring to the university (Inputs) and what they achieve upon leaving (Outcomes), the actual experience and effort invested during their tenure (Involvement) often remains opaque. This paper addresses this gap by presenting a decision-making framework grounded in established student development theories.

Theoretical Framework

The proposed framework is built upon two pillars of educational theory:

  1. Alexander Astin’s I-E-O Model: This model categorizes student data into three distinct phases: Inputs (demographics, prior academic background), Environment (the institutional setting, programs, and faculty), and Outcomes (grades, graduation, and career placement).
  2. Student Involvement Theory: This theory posits that student success is directly tied to the "quantity and quality of physical and psychological energy" that students invest in their educational experience.

By merging these theories, the authors create an "Improved Model of Student Development" that explicitly accounts for student involvement as the bridge between initial inputs and final outcomes.

Methodology

The research framework consists of three integrated systems: the Student Development System (data organization), the Educational Data Mining System (the analytical engine), and the Decision Process (the human-in-the-loop application).

The EDM process itself is designed as an iterative cycle, ensuring that data analysis is not a one-time event but a continuous improvement loop.

The iterative Educational Data Mining process framework.

As shown in the figure above, the process begins with hypothesis formation based on the educational environment. Raw data is then preprocessed and transformed into a modified format suitable for mining. Using algorithms like C4.5/J48, patterns and models are extracted, which are then interpreted and evaluated to refine the initial hypotheses and inform institutional decisions.

Case Study and Results

To validate the framework, the authors conducted a case study using a dataset of 2,266 master’s students from a Chinese university. The study aimed to predict "Job Type" (e.g., state-owned enterprise, scientific institution) based on 11 attributes covering Inputs (gender, department, origin), Involvement (average scores, prizes), and Outcomes.

The analysis was performed using the Weka data mining suite. The C4.5 decision tree algorithm achieved a predictive accuracy of 45.2%. While this numerical accuracy is relatively low, the primary value of the model lies in its interpretability. The decision tree generated specific, actionable rules for administrators. For instance, the "School" (department) emerged as the root node of the tree, indicating that departmental culture and specific curricula are the strongest predictors of career paths, even more so than individual demographics or geographic origin.

Specific rules identified trends such as:

  • Students in certain technical departments (e.g., School S5) showed a strong preference for scientific institutions.
  • Male students from specific regions looking for work in Central China were highly likely to enter research-oriented roles.

Discussion and Conclusion

The framework successfully demonstrates how EDM can be moved from a purely technical exercise to a strategic tool for university management. By shifting the focus from a "Black Box" to a "White Box" model—where student involvement is explicitly measured—administrators can gain a clearer understanding of the mechanisms driving student success.

The study's limitations, particularly the 45.2% accuracy rate, highlight the need for richer "Involvement" data. Future iterations of this work should incorporate more granular metrics, such as student-faculty interaction frequency and participation in extracurricular activities. Despite these limitations, the framework provides a robust foundation for iterative, data-driven decision-making in higher education, allowing institutions to continuously refine their environments to better support student development.

Summary updated

Abstract

Applying Educational Data Mining (EDM) for decision making is an emerging interdisciplinary research field. From the view of student development, this paper presents a framework for educational decision making, which can explore some laws and characteristics in student development to improve educational decision and quality. The paper also presents a case study to verify the effectiveness of the framework.
BibTeX
@misc{lei2017educational,
  author = {Xiaofeng Lei and Ming Yang and Yi Cai},
  title = {{Educational data mining for decision-making: a framework based on student development theory}},
  year = {2017},
  doi = {10.2991/eeeis-16.2017.76},
}

Loading related content...

Figure