datamllab / tods

TODS: An Automated Time-series Outlier Detection System
http://tods-doc.github.io
Apache License 2.0
1.47k stars 193 forks source link

Primitives for Random Oversampling or Undersampling of Data #77

Open sagarshankar opened 2 years ago

sagarshankar commented 2 years ago

Do any of the TODS primitives or any of the primitives within the d3m infrastructure perform Random Sampling like Undersampling or Oversampling? I have been trying to incorporate those steps within a pipeline prior to sending the data into the detection algorithm?

lhenry15 commented 2 years ago

Nope. At this point, we are also figuring out how to do that and see if another module is needed. Because oversampling for time series data is very different from oversampling for tabular/image data. There is a survey paper working on this. But how to incorporate that into our setting need some more investigations. Feel free to throw some ideas here. https://arxiv.org/pdf/2002.12478.pdf