Implement a sklearn style fit and predict to use SOS as one-class classifier
opengeostat opened this issue · comments
This is an elegant and intuitive algorithm for outlier detection. It would be nice to extend it to a one-class classifier by implementing a fit and predict functions in sklearn style. To do that it will be required:
- moving the content of
SOS.predict
toSOS.fit
.SOS.fit
may generate a model with matricesD
,A
,B
,O
- implementing a new
SOS.predict
with a new data pointx'
. For example, by mapping\rankingx'
on the matrixD
, or any other way that you probably already have in mind!