| Popis: |
The phrase "anomaly detection" is often used to refer to any technique that seeks to find samples that differ from expected patterns. Depending on the availability of data labels, the types of abnormalities, and the applications, many anomaly detection models are developed. This study aims to give a well-organized and thorough review of anomaly detection research. We think it will aid in a superior thoughtful of the various areas in which study has been conducted on this issue, as well as how approaches created in one field can be utilized in domains where they were not originally intended. We've divided the anomaly detection research methodologies into distinct categories. We describe the fundamental anomaly detection approach, as well as its modifications and important assumptions, for distinguishing between normal and abnormal behavior in each category. In addition, we highlight the merits and limits of each category, as well as examine the computational complexity of the approaches in real-world application areas. Finally, we discuss research gaps and limitations encountered when using deep anomaly detection algorithms to solve real-world problems. |
