Advancements in Anomaly Identification: A Comprehensive Literature Survey on Fuzzy Aggregation-based Deep Learning Approaches
Vijendra Pratap Singh, Anita G. Khandizod, Dharavath Nagesh · International Journal of Scientific Methods in Engineering and Management · 2023
Real-time monitoring systems are increasingly crucial since urbanization and industrialization have grown significantly in recent decades. AI-based anomaly detection ignores the notion that anomalous behavior’s features may change over time, making it worthless. The necessity for a typical, fixed-error training sample is another limitation of anomaly detection algorithms. This research suggests utilizing the Step Incremental Learner (SIL) to find abnormalities in live video streams. SIL uses active learning and fuzzy aggregation for unsupervised deep learning. This allows real-time updates and baseline and outlier detection. SIL is assessed on three benchmark datasets using accuracy, robustness, processing overhead, and contextual factors. Our lab experiments demonstrate the proposed device can maintain video monitoring.