Feature-weighted-FKNN-regression-using-fuzzy-mutual-information-and-Lukasiewicz-similarity

This is a new method to the family of fuzzy k-nearest neighbor (FKNN) regression based on the use of feature weights and Minkowski distance. This method is called feature-weighted Minkowski distance-based fuzzy $k$-nearest neighbor regression method (FWMD-FKNNreg).

Matlab functions:

The main functions include the the FWMD-FKNNreg algorithm (mink_weighted_fknnreg.m) and feature weights compuation (generate_feature_weights.m). In addition to those files, an example (example_run.m) of the use of FWMD-FKNNreg is also presented. FHjoin.m,MC.m, simL, and simR are needed to compute feature weights based on relevance, redundancy, and dependency.

Reference: [Kumbure, M. M., Luukka, P., Collan, M.: Feature-weighted FKNN regression using fuzzy mutual information and Łukasiewicz similarity. In: Proceeding of the IEEE International Conference on Fuzzy Systems (FUZZ-IEEE-2026)]

Created by Mahinda Mailagaha Kumbure & Pasi Luukka 03/2026