imperial-qore
Better results without normalization
https://github.com/imperial-qore/TranAD/blob/c7d8befefeb609ba823f5522b53a8262d00c2948/preprocess.py#L147
Actually, the additional normalization in SMAP dataset is not necessary as it is already normalized (as mentioned here: https://github.com/khundman/telemanom#raw-experiment-data). Moreover, the results are better when your normalization is turned off:
SMAP {'FN': 0, 'FP': 164, 'Hit@100%': 1.0, 'Hit@150%': 1.0, 'NDCG@100%': 1.0, 'NDCG@150%': 1.0, 'ROC/AUC': 0.9894289029263891, 'TN': 7593, 'TP': 748, 'f1': 0.9011998572488583, 'precision': 0.8201754296033396, 'recall': 0.9999999866310163, 'threshold': 0.20077764088704506}
Hi. Thanks for this insight. We are aware that normalization does not help in some cases; however, we have performed all preprocessing steps as per prior works to ensure fairness in our evaluation.
