Fine Grained Learning Style Prediction from Joint Analysis of Content Semantics, Knowledge Profile and Temporal Affective State

E-learning is creating a revolution in education progress with highly interactive and engaging content delivery methods.Personalized e-learning systems are one step ahead with interaction and engagement fine-tuned to learner's profile.Current systems use affective state modeling, learning style predictions to fine tune the learning process.Learning style and affective state are modeled as learner's attributes considering only content semantics and learner's knowledge profile.To improve the personalization effectiveness even more, this work proposes a finegrained learning style prediction adaptive to joint content semantics, knowledge profile and temporal affective state.The proposed solution extracts affective features using modified Densenet, extracts interaction features at various stages of course, extracts knowledge profile features from course progression and uses a novel fuzzy membership function to predict learning style at a fine-grained level.This fine-grained learning style prediction allows optimizing the learning process for maximum learning effectiveness and reduced lack of engagement.Through performance analysis, the proposed solution is found to predict learning style with an accuracy of 87% for MUTLA dataset and 98.63% for GRL-LS dataset.