Hybrid Genetic Model for Clustering Ensemble

• Impact Factor:8.139
• DOI number:10.1016/j.knosys.2021.107457
• Affiliation of Author(s):西南交通大学
• Journal:Knowledge-Based Systems
• Place of Publication:NETHERLANDS
• Key Words:ClusteringClustering ensembleGenetic algorithmHybrid genetic model
• Abstract:Clustering ensemble has received considerable research interest and led to a proliferation of studies, since it has great capabilities to combine multiple base clusters to generate a more robust and stable consensus result. Genetic algorithms are optimization methods which can search heuristically by simulating the natural evolution process with highly parallel and adaptive characteristics. However, to our knowledge, there are very few existing methods using a genetic model to solve clustering ensemble problems. In this paper, a novel hybrid genetic model for clustering ensemble (HGMCE) is proposed innovatively, and the corresponding objective function is designed. Each base clustering is regarded as a new attribute of data, and the result of clustering ensemble can be evaluated by the objective function. Then the proposed model can be inferred with the optimization, combination, and transcendence of base clustering results step by step, which makes it possible to maintain the diversity of the population and provides more possibilities to avoid falling into the local optimal solution. Furthermore, an algorithm named HGCEA corresponding to the proposed model is designed to solve the clustering ensemble problem. To evaluate the potential of HGCEA, extensive experiments are carried out on ten real datasets, including comparison with clustering groups and clustering ensemble groups. The results of accuracy and normalized mutual information demonstrate the superiority of the proposed algorithm in integrating effective clustering over the state-of-the-art.
• Co-author:YinghuiZhang,PingDenga,TianruiLi
• First Author:Wenlu Yang
• Indexed by:Academic papers
• Correspondence Author:Hongjun Wang
• Document Code:20213710877787
• Discipline:Engineering
• First-Level Discipline:Computer Science and Technology
• Volume:Volume 231
• Issue:107457
• Page Number:1-12
• ISSN No.:0950-7051
• Translation or Not:no
• Date of Publication:2021-09-02