Not only that but good anomaly detection is easy to mess up you have to be careful with how you amplify the anomalies in the dataset so that you don't wind up predicting garbage anyway.