Outlier detection is typically thought of as a pre-processing phase that seeks out objects in a data set that deviate from established expectations of behaviour. Data mining relies heavily on it to find new or unusual events, anomalies, evil deeds, remarkable phenomena, etc. Additionally, it is looking at categorical data sets' outlier detection. Due to the difficulties in constructing a useful similarity metric for categorical data, this challenge is particularly complex. With the use of a novel notion of holoentropy, which takes into account both entropy and total correlation, this study presents a formal definition of outliers and an optimisation model of outlier detection.

abnormalities, violent behaviour, and unusual phenomena.