Implementation of the technique described in
László et al (2018): Improving the Performance of the k Rare Class Nearest Neighbor Classifier by the Ranking of Point Patterns. In: Lecture Notes in Computer Science, vol 10833, p. 265-283
- doi: https://doi.org/10.1007/978-3-319-90050-6_15
- link: https://link.springer.com/chapter/10.1007/978-3-319-90050-6_15
- preprint: PREPRINT
pip install git+https://github.com/gykovacs/krnn_with_spatial_features# importing the dataset package from sklearn
import sklearn.datasets as sd
# import the KRNN_SF classifier
from KRNN_SF import KRNN_SF
# loading the IRIS dataset
X, y= sd.load_iris(return_X_y= True)
# turning the IRIS multi-classification problem into an unbalanced binary classification
y[y == 2]= 1
# fitting and predicting
krnn_sf= KRNN_SF(correction= 'r1')
krnn_sf.fit(X, y)
krnn_sf.predict_proba(X)