Predicting Students' Learning Styles in an Educational Digital Game System Using Support Vector Machines

Games play a vital role in human culture and society, fostering motivation and engagement. This is why the principles of gaming, such as gamification and interactive design, are being applied in various non-gaming settings like schools, healthcare, and consumer behaviour to improve motivation, behaviour, and learning outcome. This study aims to predict students' learning style preferences based on their behaviour in an educational digital game system using the Support Vector Machines. A dataset extracted from the Jowilder competition on the Kaggle platform is used to train a Support Vector Machines classifier, which identifies learning style preferences. Hyperparameter optimization is performed using kernel function tuning to enhance model accuracy. The Support Vector Machines model is trained using different kernels-Linear, Polynomial, and Radial Basis Function to verify which approach works best for classification. The assessment of the model using accuracy, precision, recall, and F1-score indicates that the Radial Basis Function kernel performs the best, achieving a high accuracy. This study reveals that gameplay behaviour can provide a reasonable predictor for learning styles, opening up new opportunities for personalized learning paths in serious gaming environments.