An Analysis of Data Anomalies in Data Mining and Knowledge Discovery in Data

The viewpoint from which data anomalies are defined, detected and managed may vary based on a given data set and the specific goals of an application. Anomalous data can be viewed as extraneous noise that negatively impacts a data mining analysis. Alternatively, discovering data anomalies may be the actual goal of an application. The authors cite several definitions for data anomalies and present a review of some data anomaly types based on a survey of current literature. They review and contrast examples of how the same anomalies are defined in different taxonomies and give examples of methods used for detecting them in data sets, and describe some ways that they can be managed.