Construction and application of evaluation models for online learning behaviour

Abstract

Online learning not only makes full use of online resources to meet the diverse learning needs of students but also enhances their autonomy and information acquisition abilities. Through learning analysis techniques, the quality of online learning can be improved, thereby enabling evaluation, diagnosis, prediction, and intervention in learning. Based on the Ulearning platform, this study used 792 undergraduate course score data from 2023, divided students’ online learning behaviours into different indicators, and performed exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) on the students’ online learning data to verify the feasibility of the learning evaluation model indicators. Then, through the demand analysis of teaching decisions, the dataset was split by K-fold cross-validation, a nonlinear support vector machine (SVM) was used to optimize the hyperparameters to establish an online learning behaviour evaluation model. The experimental results show that factor analysis effectively extracts three core factors (Platform usage intensity, Learning behavior, Effective learning intensity) affecting academic performance and greatly reduces data dimensionality. The proposed model achieves a high coefficient of determination (R2) of 0.960 and a low root mean square error (RMSE) of 0.194, demonstrating outstanding predictive performance and strong generalization ability. This method provides a reference basis for further improving students’ online learning effectiveness and can be extended to student modeling in other types of learning activities. More importantly, it can capture the interdependence between different aspects of students’ learning experiences and learning outcomes, thereby guiding teachers to better carry out teaching work, make early interventions, and allocate resources more effectively.