A Non-symmetrical Weighted K-means with Rank-Based Artificial Bee Colony Algorithm for Medical Diagnosis
Jenn-Long Liu, Chung-Chih Li · International Journal of Machine Learning and Computing · 2015
This study proposes a non-symmetrical weighted k-means (NSWKM) clustering algorithm to improve the accuracy of clustering result.The similarity distance of original k-means algorithm is modified by adding weights with a non-symmetrical form to the distance measurement.Namely, different weights for attributes are applied for clusters such that the contribution of attributes can be adjusted adaptively during the clustering process.In this work, the weights are given via an optimization process using a rank-based artificial bee colony (RABC) algorithm.Furthermore, the proposed NSWKM clustering algorithm combined with the RABC, termed NSWKM-RABC herein, is then applied to the medical diagnoses of five data sets of diseases, including breast cancer, cardiac disease, diabetes, liver disease and hepatitis, to evaluate the performance of the proposed algorithm. Index Terms-Data sets of diseases, medical diagnoses, non-symmetrical weighted k-means clustering algorithm, rank-based artificial bee colony algorithm.I.