Clustering-based rule generation methods for fuzzy classifier using Autonomous Data Partitioning algorithm
Светлаков Михаил Олегович, I. A. Hodashinsky · Journal of Physics Conference Series · 2021
Abstract In this paper, clustering-based rule generation methods for fuzzy classifier using non-parametric Autonomous Data Partitioning algorithm have been proposed. ADP-algorithm is used to determine the number of clusters for use in various k-means-like clustering algorithms. Proposed method contributes to solving the problem of determining optimal number of clusters/rules. The efficiency of fuzzy classifiers with rules constructed by the specified algorithms has been tested on data sets from the KEEL repository. Experimental results show that proposed method outperforms baseline algorithm (the extremums rulebase generation algorithm) both in terms of classification accuracy and geometric mean metrics.