Hybrid LSTM-Based Traffic Anomaly Detection for Smart Mobility in Metaverse Cities

The rapid development of Metaverse technologies has created new opportunities for modeling and managing intelligent transportation systems in virtual urban environments. However, ensuring efficient and stable mobility within these digital ecosystems requires accurate and interpretable anomaly detection mechanisms. This study proposes a Hybrid LSTM–Isolation Forest (HLIF-Net) framework for identifying traffic anomalies in smart mobility simulations using the METR-LA dataset. The model integrates a deep learning-based Long Short-Term Memory (LSTM) network for sequential traffic prediction with an Isolation Forest algorithm that detects anomalies from residual prediction errors. The proposed framework was trained using twelve-step input sequences of normalized traffic speed data and evaluated across 34,260-time samples. Experimental results demonstrated strong model stability, with a Mean Squared Error (MSE) of 0.0021 on training data and 0.0025 on validation data. Approximately 3 percent of traffic instances were classified as anomalous, reflecting potential irregularities such as congestion, sudden speed changes, or sensor inconsistencies. Temporal and spatial analyses further revealed that anomalies tend to cluster during periods of instability and concentrate in high-mobility regions. These findings confirm that the HLIF-Net framework provides a robust, data-driven solution for real-time anomaly detection and intelligent mobility management in Metaverse city environments.