Virtual Reality (VR) is transforming various domains by providing immersive experiences, but a significant barrier still exists in the inability to recognize user immersion and involvement precisely. To address this study introduces the Adaptive Virtual Reality Design using Immersion Levels and User Engagement (AVRDIL-UEGE) algorithm, which leverages an ensemble of deep learning models to optimize VR immersion and user engagement. By employing a multi-stage approach, the model first standardizes input data using Linear Scaling Normalization (LSN) to enhance accuracy. Subsequently, it integrates Stacked Long Short-Term Memory (SLSTM), Stacked Sparse Autoencoder (SSAE), and Elman Neural Network (ENN) in an ensemble framework to capture complex temporal and spatial patterns. Hyperparameter tuning is performed using the Multi-strategy Improved Red Deer Optimization (MIRDO) algorithm, ensuring robust model performance. Experimental results on the benchmark VR dataset demonstrate the superior predictive accuracy of AVRDIL-UEGE, with a classification accuracy of 98.40% and an MCC of 0.95 compared to traditional machine learning approaches. The proposed model offers significant improvements in user immersion detection, paving the way for more responsive and personalized VR experiences. This work contributes to advancing VR analytics by providing a data-driven, scalable solution for immersive design optimization in real-time applications.
Science
