zuoan123/ada-knn

A MATLAB implementation of Adaptive k-Nearest Neighbor Algorithms called Ada-kNN and Ada-kNN2 (alongside a global weighting scheme for handling class imbalance).

A MATLAB implementation of Ada-kNN, Ada-kNN2, Ada-kNN+GIHS and Ada-kNN2+GIHS.

Written by: Sankha Subhra Mullick.

Reference: S. S. Mullick, S. Datta and S. Das, "Adaptive Learning-Based k-Nearest Neighbor Classifiers With Resilience to Class Imbalance," in IEEE Transactions on Neural Networks and Learning Systems, doi: 10.1109/TNNLS.2018.2812279.

Contact: [email protected] (Sankha Subhra Mullick).

DESCRIPTION:

• The package contains 7 functions.
• adaKnn.m: Function implementing Ada-kNN algorithm.
• adaKnn2.m: Function implementing Ada-kNN2 algorithm.
• adaKnnGIHS.m: Function implementing Ada-kNN coupled with GIHS algorithm (for imbalanced classification).
• adaKnn2GIHS.m: Function implementing Ada-kNN coupled with GIHS algorithm (for imbalanced classification).
• kNNIMB.m: Function implementing weighted k-nearest neighbor algorithm.
• learningModel.m: Function implementing the proposed heuristic learning technique (to be used by adaKnn2 and adaKnn2GIHS).
• standardised.m: Supporting function.

DEPENDENCIES:

• MATLAB 2014a and above.
• Neural Network toolbox (for adaKnn and adaKnnGIHS).