A novel framework to generate clustering algorithms based on a particular classification structure
Hossein Karami, Mohammad Taheri · 2017
Classification and clustering are two main tasks of pattern recognition. Ensemble of classifiers or clustering algorithms is one of the ways to provide a robust, accurate and stable final result. In addition, clustering may be used to improve the performance of a classifier or vice versa. In this paper, a novel framework is proposed as an ensemble of classification and clustering algorithms. In this framework, clustering can be done based on the structure of a base classifier. By use of this framework, new clustering methods can be generated, or some classic ones may be regenerated considering underlying theory of a particular classifier. As a sample of the proposed framework, Parzen windows classifier is used as the base classifier to generate a variety of clustering algorithms including some well-known methods, complete and single linkage.