“…Unsupervised methods of outlier detection have been studied widely because they require no additional labels. Representative algorithms include proximity- [3,7,36,44], statistics- [53,61], cluster- [21,33,47,55], OC-SVM- [13,14,54], and reconstruction-based [32,45,59, 60] models. Proximity-based models assume that outliers are points that are far from the other data, and can be identiied by measuring their distance or density.…”