christophraab / rrslvq

Code release of Reactive Robust Learning Vector Quantization

Reactive Soft Prototype Computing for frequent reoccurring Concept Drift

Python source code of the publication Christoph Raab, Moritz Heusinger, Frank-Michael Schleif, Reactive Soft Prototype Computing for Concept Drift Streams, Neurocomputing, 2020.

The repository contains the Kolmogorov-Smirnov Windowing (KSWIN) concept drift detector and the Reactive Robust Soft Learning Vector Quantization (RRSLVQ), but the later is the main contribution of the paper.

The RRSLVQ is made for classifying streams with very high rates of drift while maintaining stabilit during active concept drift

Demo of RRSLVQ and a demonstration of stability during drift is available at

demo.py

Requirements: scikit-multiflow:https://github.com/scikit-multiflow/scikit-multiflow bix: https://github.com/ChristophRaab/bix

rrslvq's Issues

problem to run the code

when i try to run the demo i get
Failed to import SciKit-Multiflow!
Traceback (most recent call last):
File "rrslvq.py", line 58, in
class RRSLVQ(ClassifierMixin, skmultiflow.core.base.StreamModel, BaseEstimator):
NameError: name 'skmultiflow' is not defined

the problem is that i can not run the following command
from skmultiflow.core.base import StreamModel

not being able to run it on my windows PC

Traceback (most recent call last):
File "rrslvq.py", line 516, in
stream = ReoccuringDriftStream(stream=s1, drift_stream=s2,random_state=None,alpha=90.0, position=2000,width=1)
TypeError: Can't instantiate abstract class ReoccuringDriftStream with abstract methods prepare_for_use

Running `study_perfomance_detectors.py` raises `ModuleNotFoundError`

I'm trying to run the study_perfomance_detectors.py but the error ModuleNotFoundError is raised due the file cd_naive_bayes.py not being in the study folder.

I wondering how you run this code. I know that if I move the files from one folder to another I can solve the error, but if there is a simpler and/or cleaner way to solve I would like to learn.

Distinguishing abrupt and gradual drift

I run an implementation of your work based on skmultiflow implementation here:
https://scikit-multiflow.readthedocs.io/en/stable/api/generated/skmultiflow.drift_detection.KSWIN.html

But I couldn't find any way to distinguish between abrupt and gradual drift that you discuss in the paper. Could you please explain whether this feature is implemented or not? Thanks.

Can i get bix.detectors.ksvec source?

Hi, can i get the source of bix.detectors.ksvec in https://github.com/ChristophRaab/rrslvq/blob/master/cd_naive_bayes.py?