Abstract:

The massive increase in data stored in the educational domain, on the one hand, and the increased power of available software tools and machine algorithms for data analysis, on the other, have resulted in the development of the fields of educational data mining (EDM) and learning analytics (LA). Data mining employs a vast knowledge base, advanced analytical skills, and domain knowledge to uncover hidden trends and patterns that can be applied in almost any domain, from business to education. Educational institutions can use data mining to extract useful information from their databases. Institutions of higher education frequently use learning management systems (LMS) to support the teaching and learning process. The most widely used LMS in Croatia, specifically at the University of Zagreb, is Moodle. Each student's activity on Moodle is recorded in a log file when they use their personal account. Such data is valuable source for examination. The main objective of LA is to analyze raw data in log files in order to generate new knowledge about student behavior. In this research paper, student logs from educational databases will be analyzed and predictive models to forecast students’ performance (grade/engagement) will be created using different machine-learning algorithms. Various machine-learning approaches will be employed to detect which one provied the best results on educational databases. Based on the predictive models, sensitivity analysis will be performed. The findings of the study revealed students behavioral patterns at the LMS and identified student activities at the LMS that results in effective course completion. These findings provide benefits to both, teachers and students. To teachers can serve as guidelines for effective course design, whereas to students can help in optimizing LMS activity. Machine learning application can enhance learning outcomes through data mining and predictive learning analytics. Findings from this research study will be used for early detection of students experiencing difficulties in a course and for course modification.

Keywords:

Predictive modeling, educational data mining, learning analytics, educational database.