Institutes outside Greece
Models and algorithms for performance prediction and course recommendation in higher education
Abstract
dc:descriptionEducational institutions should use supportive tools to reduce high student dropout rates and ensure their timely graduation. Educational data mining involves the development of such methods that leverage student data. Their purpose is to generate alerts for students at risk of failure, identify enrollment practices that are associated with academic success, and enable appropriate interventions. Applying these models to a higher education institution can improve advising and degree planning. The aim of this thesis is to develop methods that will support students in making informed decisions about their enrollment in courses within the context of higher education. We also investigate equity issues that may arise in a course recommendation system. We want to create models that can be used before the semester begins, in order to allow students to make any necessary adjustments to their plans. Instructors can also benefit from them, as it will allow them to form their expectations about the students enrolled in their class and adjust the curriculum and course content accordingly. For the purposes of this paper, we will focus on traditional higher education datasets and develop models that do not require any information about the structure and curriculum of a particular course. All we use is the data of students’ analytical scores.In this context, we first introduce next-semester grade prediction methods to estimate the grades that a student is expected to receive for the courses he or she plans to take in the following semester. These algorithms are based on sparse linear and low-order matrix factorization models that are specific to each course or student-course tuple. These methods identify predictive subsets of past courses on a course-by-course basis. Second, we delve into the factors that influence student performance. We extract informative features from the student analytic score data and formulate the problem as a binary classification problem to identify students who are performing poorly. We examine a student’s performance in both an absolute sense (in terms of grades achieved) and a relative sense (in comparison to the student’s previous performance). We present a comprehensive study to answer the following questions: which features are good indicators of a student’s performance? Which features are most important? The findings are interesting, as different features are most important for different classification tasks. Next, we address the problem of course recommendation. Recommender systems have been widely used in various fields to support decision-making and empower users’ choices. In the education sector, recommendations can generate a personalized course list for each student considering taking the next semester. To this end, we propose Scholars Walk, a random walk-based approach that captures the sequential relationships between different courses. Based on the "wisdom of crowds" and students’ previous courses, we recommend a short course list for the next semester. Our framework is highly efficient, easy to interpret, and able to take into account important aspects of the education sector. Finally, we explore fairness in the context of course recommendation. It is important to ensure that the models we develop and implement in an educational institution do not discriminate against specific groups of students. All students should be treated fairly and offered similar opportunities and quality of services based on the concept of group justice. Based on this idea, we examine the fairness of the opportunities offered to students by the system's recommendations. We formulate our approach as a multi-objective optimization problem and study the trade-offs between equal opportunities and quality. The algorithm first makes an initial assignment of courses to recommend to students, which we assume are of optimal quality. The model then improves the initial solution to improve overall fairness. Experimental evaluation with synthetic and real datasets demonstrates that we can promote equal opportunities in recommendations without significantly weakening their quality.
Degree
thesis:*- Grantor dc:publisher
- Institutes outside Greece
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Polyzou, Agoritsa
- Πολύζου, Αγορίτσα
Subjects
dc:subject × 17- Τεχνητή νοημοσύνη στην εκπαίδευση
- Αριθμητικά μοντέλα
- Σύστημα σύστασης
- Προγνωστικά συστήματα βαθμολογίας
- Προγνωστικά μοντέλα
- Predictive models
- Recommendation systems
- Educational features
- Fairness
- Higher education
- Student performance prediction
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Τεχνολογία μέσων
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Media Technology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/60070