AutoML Anomaly Characterization Method

As datasets become increasingly complex, traditional anomaly detection methods often fall short, necessitating more sophisticated techniques. This paper introduces a novel AutoML method for identifying and classifying a wide range of data anomalies, in complex datasets. This approach effectively manages null anomalies and deviations from standard deviations through robust statistical analysis and the intentional introduction of text and null errors in numeric features. Tested on five diverse datasets, the method demonstrated flawless performance in identifying a high percentage of induced errors and accurately distinguishing between different types of data features and anomalies. This preliminary study highlights the method's potential for further application and validation on datasets with real-world anomalies and untested anomaly types, aiming to enhance data quality assessment both before and after processing.