<![CDATA[Student learning interest is an important psychological factor associated with student engagement and learning experiences in Physical Education (PE). However, limited research has examined whether observable behavioural indicators, particularly attendance and classroom participation, can be integrated into predictive modelling of students’ learning interest and whether deep learning provides better predictive performance than conventional statistical approaches. This study aimed to examine the relationships of student attendance and classroom participation with learning interest in PE and to evaluate the predictive performance of a deep learning model using these behavioural indicators. A quantitative correlational-predictive research design was employed involving 150 senior high school students from North Luwu and East Luwu Regencies, Indonesia, selected through stratified random sampling. Data were collected using validated learning interest questionnaires, school attendance records, and structured classroom observations of student participation. Descriptive and structural model analyses were conducted to examine the relationships among the study constructs, followed by deep learning modelling to evaluate predictive performance on an independent test set. The findings indicated that student attendance patterns and classroom participation were significantly and positively associated with learning interest in PE. The structural model explained a substantial proportion of the variance in learning interest (R² = 0.904). In the independent test-set evaluation, the deep learning model achieved lower prediction errors (MAE = 0.194; RMSE = 0.267) and a higher R² (0.887) than the conventional statistical model (MAE = 0.286; RMSE = 0.354; R² = 0.821). These findings indicate that the deep learning model provided better predictive performance under the evaluation procedure used in this study. The study contributes to educational analytics by demonstrating how observable behavioural indicators can support predictive modelling of learning interest in PE. The findings also have potential implications for Islamic Education, where behavioural indicators such as attendance and active participation may complement teachers’ professional judgment in supporting student engagement, discipline, and character-oriented learning.]]>
Copyrights © 2026
