A real-time AI tool for hybrid learning recommendation in education: Preliminary results

In recent years, owing to previous pandemics, online and offline education have collaboratively influenced students' academic experiences globally. This mixed-methods strategy offers both benefits and drawbacks for students. This study created an innovative AI tool utilizing the Support Vector Machine (SVM) algorithm on primary samples of Hungarian informatics students to assess their suitability for adopting hybrid learning in their studies. This paper also explained the strength of the model with Shapley Additive exPlanations (SHAP) values of the Explainable Artificial Intelligence (XAI) method. The SVM parameters were suggested by Grid Search Cross Validation (GSCV) to provide an optimal classifier with maximum accuracy. The trained model has been tested and validated with new realistic samples created with a non-parametric approach, Kernel Density Estimation (KDE). The newly created test data sets correspond to the actual data distribution of the data set and preserve the relationships between classes of characteristics, as validated by the Kolmogorov–Smirnov (KS) test (p < 0.05). The average precision cross-validation of training and testing the data set is 0.9881 and 0.9757, respectively, for classifying the target class of the hybrid learning data set. Additionally, the model achieved the highest f1-score, precision, and recall with values of 0.98. The Area Under Curve (AUC) is calculated to be 1, indicating a high value. According to the SVM coefficient, Overall Satisfaction (OS), Long Term Solution (LRS), and Feeling Happiness (FH) were the most influential attributes, with positive coefficients of 0.78, 0.49, 0.47 and 0.41 respectively. The negative coefficients for the Challenging Group task (CGT), and Obstacle (OBST), are -0.42, and -0.16, respectively. The SHAP explained the SVM model strength and voted for OS, FH, LRS, and CGT in the decision. The study's findings can help students and institutional administration decide whether to switch to hybrid education based on pros, cons, and other recommended features.