Anomaly Detection using Deep learning

Abstract

One of the primary goals of video surveillance studies and applications is the identification of anomalous behaviour. Public spaces including streets, junctions, banks, retail malls, etc. are progressively installing surveillance cameras to deter crime and protect their patrons. Accident, crime, and unlawful activity detection are all vital functions of video surveillance. In comparison to everyday life, unusual occurrences are very uncommon. In order to be useful, an anomaly detection system must be able to promptly alert users to any behaviour that deviates from typical patterns and pinpoint when exactly that behaviour took place. Anomaly detection may, therefore, be seen as a crude level of visual comprehension that serves to separate out abnormalities from regular patterns. Once an outlier has been identified, classification methods may assign it to a certain action. Within the framework of banking operations, this study provides an overview of anomaly detection. Multiple parties, including workers, clients, debtors, and external entities, participate in or are affected by the various activities and transactions that make up banking operations on a daily, monthly, and a periodic basis. Although certain events may take time to develop, their negative consequences may often be mitigated or even avoided altogether if they are seen early enough. Anomaly detection applied to time series is used to identify unauthorised visitors. To distinguish between typical and unusual occurrences, this study employs a machine learning-based anomaly detection approach.