An improved hybrid ontology-based approach for online learning resource recommendations

In recent years, online learning has become more and more popular. However, because of information overload, learners often find it difficult to retrieve suitable learning resources. Although many scholars have proposed excellent online learning resource recommendation algorithms, the accuracy of pe...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Educational technology research and development 2021-10, Vol.69 (5), p.2637-2661
Hauptverfasser: Shanshan, Shang, Mingjin, Gao, Lijuan, Luo
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In recent years, online learning has become more and more popular. However, because of information overload, learners often find it difficult to retrieve suitable learning resources. Although many scholars have proposed excellent online learning resource recommendation algorithms, the accuracy of personalized recommendation results still needs to be improved. This study proposes an improved hybrid ontology-based approach for online learning resource recommendations, combining collaborative filtering algorithm and sequential pattern mining (SPM) techniques. Ontology can be used effectively for knowledge representation to avoid cold start and data sparsity problems. And the history of learners’ sequential access patterns helps in providing recommendations that are more consistent with the law of learning activities. Experimental results reveal that our improved hybrid approach for learning resource recommendations yields better performance and recommendation quality than other related algorithms. Compared with previous research outcomes, our collaborative filtering engine, with ontology domain knowledge, makes full use of the historical learning paths of similar learners. The ontology construction in this study has a more reliable theoretical basis and the selection of features is more representative. In addition, improvement of the SPM process further improves the efficiency of our recommended algorithm.
ISSN:1042-1629
1556-6501
DOI:10.1007/s11423-021-10029-0