Drift Type Classification
Distribution Based
- Virtual Drift
- New Target
- Real Drift
Time Based
- Sudden Drift
- Incremental Drift
- Reoccuring Drift
- Gradual Drift
Drift Detection Mechanisms
- Drift Detection Method (DDM, EDDM, RDDM)
- Window and Statistical Based Mechanisms
- ADWIM
- Kolmogorov Smirnof Test
- Unsupervised or semi supervised methods
- Ensemble Methods
- Neural Networks
Comparison of Methods
| Method | Accuracy | Computational Cost | Applicability |
|---|---|---|---|
| DDM | High | Low | Cost Effective and easy to apply in real time |
| WBM | Medium | Medium | better than DDM for gradual drift, easy to apply in real time |
| USSM | Medium | Medium | Works best with novel class detection |
| EM | Very High | High | can be applied for various data types but costly |
| NN | Very High | High | Mostly have higher cost but higher accuracy |