Article
DEEP LEARNING-BASED USER BEHAVIOR CLASSIFICATION TO IMPROVE VIRTUAL REALITY INTERACTION
This thesis introduces a Virtual Reality (VR) User Experience Classification framework powered by deep learning, aimed at improving the design quality and user satisfaction of VR systems. The approach utilizes a rich dataset consisting of over 10,000 user interaction instances sourced from various VR environments. A key challenge identified in the dataset is class imbalance, where nearly 60% of the data belongs to the majority class, and the remaining 40% is distributed among minority classes. To address this, the study proposes a hybrid method that integrates the Synthetic Minority Over-sampling Technique (SMOTE) with a Deep Neural Network (DNN) classifier. SMOTE is employed to synthetically generate new instances for the minority classes, thereby balancing the dataset without compromising original data integrity. Subsequently, the DNN model learns hierarchical and abstract features from the balanced dataset, enabling it to capture complex behavioral patterns and interactions effectively. The proposed SMOTE-DNN model was evaluated using a traintest split approach and benchmarked against traditional baseline models. The experimental outcomes show a significant enhancement in performance, achieving over 99% across accuracy, precision, recall, and F1 score. These results affirm the efficacy of the proposed hybrid model in classifying VR user experiences and offer valuable insights for future improvements in VR system design and personalization
Full Text Attachment
