University of Patras
Ανάπτυξη πρωτότυπων αλγορίθμων μηχανικής μάθησης για χρήση σε εκπαιδευτικά δεδομένα και σε συστήματα διαχείρισης εκπαιδευτικού περιεχομένου
Abstract
dc:descriptionThe use of information and communication technologies in the educational field, contributes daily to the production and storage of large amounts of data. The need for an effective analysis of this data for retrieving valuable information and their converting into systematic knowledge resulted in the genesis of the fields of Educational Data Mining and Learning Analytics. Educational Data Mining focuses primarily on the development and implementation of data mining methods on educational data for solving important educational problems, while Learning Analytics is centered mainly around the learning process, exploiting data analysis in order to reinforce the decision-making procedures. However, irrespectively of the method used to approach an educational problem, both scientific fields have common goals; the improvement of learning and enhancing the quality of education offered. The prognosis of students’ learning outcomes constitutes one of the most significant problems in the fields of Educational Data Mining and Learning Analytics. The handling of this specific problem relates to the creation of a classification or regression model, depending on the nature of the outcome variable, implementing a supervised learning algorithm, such as a decision tree. However, the building of a predictive learning model presupposes the training of the algorithm in a set of labeled data. The difficulty of acquiring labeled data has resulted to the development of new machine learning methods which are generally referred to as “Weakly Supervised Learning”. Semi-Supervised learning and Active Learning constitute the main components of Weakly Supervised Learning with a view to exploiting a small number of labeled examples together with a large number of unlabeled ones in the best possible manner, for building accurate and robust learning models. In recent years, a plethora of Semi-Supervised Learning algorithms have been developed and implemented with great success for solving a variety of problems in many scientific fields. However, the effectiveness of these methods has not been thoroughly studied in the educational field, as it is evident from the pertinent literature, generating, thus, new challenges for scientists and researchers. These challenges relate not only to the implementation of existing Semi-Supervised Learning algorithms but also to the development of innovative algorithms for extracting valuable knowledge from various sources of educational data. Within this frame, the main purpose of the present thesis is the development of new Semi-Supervised Learning methods and their implementation for the prognosis of students’ learning outcomes at different levels of education. More specifically, we develop a co-training style algorithm and a semi-supervised regression algorithm for predicting the academic performance and the grades of undergraduate students in distance higher education, respectively. The proposed methods are deemed to be very effective for the accurate and early prediction of students’ learning outcomes, as confirmed by the experimental results of our published studies. Rounding up, we consider that the present thesis constitutes the first systematic and comprehensive attempt towards exploiting the Semi-Supervised Learning framework in the educational field, anticipating better results than those generated from traditional supervised methods.
Degree
thesis:*- Grantor dc:publisher
- University of Patras
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kostopoulos, Georgios
- Κωστόπουλος, Γεώργιος
Subjects
dc:subject × 24- Εξόρυξη γνώσης από εκπαιδευτικά δεδομένα
- Μαθησιακή αναλυτική
- Μηχανική μάθηση
- Ημι-επιβλεπόμενη μάθηση
- Πρόβλεψη
- Αλγόριθμοι κατηγοριοποίησης
- Αλγόριθμοι παλινδρόμησης
- Αλγόριθμος συνεκπαίδευσης
- Educational data mining
- Learning analytics
- Machine learning
- Semi-supervised learning
- Prediction
- Classification algorithms
- Regression algorithms
- Co-training algorithm
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Κοινωνικές Επιστήμες
- Εκπαίδευση
- Natural Sciences
- Computer and Information Sciences
- Social Sciences
- Education
Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/47275