A Comprehensive Study on Anomaly Detection Methods Using Traditional and Machine Learning Approaches

Detecting anomalies is critical in multiple domains such as cyber security, network security, etc.This paper provides an overview of anomaly detection methods, ranging from traditional statistical to modern ML approaches.The paper begins by explaining anomalies and their significance, followed by a detailed overview of traditional methods such as PCA and Z-Score Analysis.Then, there is a transition to methods based on machine learning that use supervised and unsupervised algorithms such as OCSVM & Bayesian Networks, followed by an illustration of the applications of these methods using examples in fraud detection.This paper also comparatively evaluates different methods, highlighting their advantages and limitations.In the end, this is a proposal for future research to address current limitations.This paper explains anomaly detection and its effective applications in various domains.