Contextual vs Structural #59
Answered
by
kayzliu
jdschmitt11
asked this question in
Q&A
-
Many your datasets have both structural and contextual outliers injected into the datasets. Is it possible to determine whether it's a 1: contextual, 2: structural, or 3: both after the model as been fit? It looks like model.label_ and model.predict() both return either 0 or 1 and don't tell us which type it is. Thank you! |
Beta Was this translation helpful? Give feedback.
Answered by
kayzliu
Nov 25, 2022
Replies: 1 comment
-
Thanks for your suggestion. The original models only predict whether the nodes are outliers or not. So current implementation won't be able to tell the type of outliers. But in the future, we may cache the intermediate score for feature and structure reconstruction, which may help determine the type of outliers. |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
jdschmitt11
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Thanks for your suggestion. The original models only predict whether the nodes are outliers or not. So current implementation won't be able to tell the type of outliers. But in the future, we may cache the intermediate score for feature and structure reconstruction, which may help determine the type of outliers.