Skip to content

unsupervised concept drift detection with one-class classifiers

Notifications You must be signed in to change notification settings

ogozuacik/one-class-drift-detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 

Repository files navigation

Concept Learning using One-Class Classifiers for Implicit Drift Detection in Evolving Data Streams

Parameters:

  • nu: parameter for SVM (set it to 0.5)
  • size: window size
  • percent: threshold for outlier percentage

Command line instructions:

  • python OCDD.py dataset_name nu size percent (sample: python OCDD.py elec.csv 0.5 100 0.3)

  • You can either put the datasets into the same directory or write dataset directory in place of dataset_name. Datasets should be in csv format. You can access the datasets used in the paper and more from:

  • You have to install scikit-multiflow in addition to commonly used python libraries. (sklearn, pandas, numpy, matplotlib)

The code will output:

  • Final accuracy
  • Total elapsed time (from beginning of the stream to the end)
  • Prequential accuracy plot (dividing data stream into 30 chunks)

About

unsupervised concept drift detection with one-class classifiers

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages