Predicting Students' Performance in Mixed-mode Learning Strategy using Neural Network Modelling

This paper predicts students' performance using neural network (NN) modelling within the Science, Technology, Engineering, and Mathematics (STEM) subject. The mixed-mode learning (MML) strategy is considered in this study based on Bloom's Taxonomy, flipped classroom, asynchronous and cognitive learning. Standard statistical analyses are included to explore the: i) relationships between the variables; ii) statistically significant predictor variables for the NN and; iii) regression analysis with estimated regression coefficients. The Broyden-Fletcher-Goldfarb-Shanno (BFGS) Quasi-Newton algorithm is applied as the training rule for NN. The prediction results provide an insight of the MML strategy in supporting and improving students' learning experience and their final academic performance. The NN model has also achieved a good level of prediction accuracy.