Break-up-anomaly detection

The definition of the anomaly refers to the hawkings outliers definition. The problems needing attention include the number of attributes, the global/local, the degree of anomaly, the number of recognition anomalies, and the evaluation. Detection methods are: model-based approach, proximity-based approach, density-based approach. Under the model-based method, the standard deviation of the one-element normal distribution is used, and the Mahalanobis distance is used to differentiate the multivariate normal distribution, which takes into account the shape of the distribution and the maximum likelihood estimation using the mixed model for the shielding (masking) problem. In the proximity-based approach, the greater the number of neighbors K, the more likely it is to be detected as an anomaly. In density-based methods, density can be calculated by inverse distance and point count, and relative density is used when the cluster density is uneven.

Break-up-anomaly detection