Naslov
Mining informatics’ courses logs to predict students’ stress

Autori
Đurđević Babić, Ivana

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
DIDACTIC CHALLENGES IV: FUTURES STUDIES IN EDUCATION Conference Proceedings / INAYATULLAH, SOHAIL ; DUBOVICKI, SNJEŽANA ; BILIĆ, ANICA - Osijek : Josip Juraj Strossmayer University of Osijek Faculty of Education, Osijek, Croatia and Croatian Academy of Sciences and Arts, The Center for Scientific Work in Vinkovci, 2023, 483-491

ISBN
978-953-8371-15-8

Skup
Didactic Challenges IV: Futures Studies in Education

Mjesto i datum
Osijek, Hrvatska, 26.05.2023. - 27.05.2023

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
data mining, informatics, log data, neural network, students

Sažetak
The current pandemic situation has caused significant changes in education and the use of learning management systems (LMS) increased with an obvious need for online communication and collaboration. Such a sudden change in the organization of teaching, evoked stress in stakeholders in education. Since it is known that stress can negatively affect students’ performance, this paper investigates whether students with higher or lower stress levels in the class average can be effectively predicted by mining logs of the informatics’ courses from the LMS. Also, the aim is to reveal which variables are important for models’ accuracy. Log data were obtained from the Loomen LMS and the perceived stress level of 126 students was collected. The results show that the mean value of participants’ perceived stress is 22.85 (SD =6.04) and that there is a weak negative statistically significant correlation between perceived stress and the page component of the LMS log files at a 5% significance level (r=-. 2). The best neural network model achieved an overall accuracy of 66.67%. It was more effective in detecting students with higher than average stress levels among participants (75%) than those with lower stress levels (60%). The file and system components had the greatest impact on the model’s performance, as revealed by the sensitivity analysis.

Izvorni jezik
Engleski

Znanstvena područja
Informacijske i komunikacijske znanosti

Napomena
Konferencija je održana u suorganizaciji s Hrvatskom
akademijom znanosti i umjetnosti (HAZU), a zbornik
konferencije i izdan u suizdavaštvu s HAZU.

POVEZANOST RADA