A comprehensive survey of traditional, merge-split and evolutionary approaches proposed for determination of cluster number

Hançer, Emrah; Karaboğa, Derviş

doi:10.1016/j.swevo.2016.06.004

Cited by 108 publications

(54 citation statements)

References 174 publications

(179 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Hancer E. and Karaboga D. [26] created a comprehensive survey of methods related to automatic cluster number evaluation, however the most similar works to our paper are the Silhouette method, which was first described by Peter J. Rousseeuw [47] and the cluster validity proposed in [52]. In this paper we compare the performance of our solution to these techniques.…”

Section: Related Workmentioning

confidence: 91%

MMKK++ algorithm for clustering heterogeneous images into an unknown number of clusters

Papp

Szücs

2018

ELCVIA

View full text Add to dashboard Cite

In this paper we present an automatic clustering procedure with the main aim to predict the number of clusters of unknown, heterogeneous images. We used the Fisher-vector for mathematical representation of the images and these vectors were considered as input data points for the clustering algorithm. We implemented a novel variant of K-means, the kernel K-means++, furthermore the min-max kernel K-means plusplus (MMKK++) as clustering method. The proposed approach examines some candidate cluster numbers and determines the strength of the clustering to estimate how well the data fit into clusters, as well as the law of large numbers was used in order to choose the optimal cluster size. We conducted experiments on four image sets to demonstrate the efficiency of our solution. The first two image sets are subsets of different popular collections; the third is their union; the fourth is the complete Caltech101 image set. The result showed that our approach was able to give a better estimation for the number of clusters than the competitor methods. Furthermore, we defined two new metrics for evaluation of predicting the appropriate cluster number, which are capable of measuring the goodness in a more sophisticated way, instead of binary evaluation.

show abstract

Section: Related Workmentioning

confidence: 91%

MMKK++ algorithm for clustering heterogeneous images into an unknown number of clusters

Papp

Szücs

2018

ELCVIA

View full text Add to dashboard Cite

show abstract

“…Kümeleme bir veriyi küme içi benzerliğin maksimum ve kümeler arası benzerliğin minimum olacak şekilde alt kümelere (gruplara) ayırma işlemidir. Kümeleme algoritmaları sistematik olarak hiyerarşik, bölünmeli, yoğunluk tabanlı, alan tabanlı ve model tabanlı olmak üzere beş grupta incelenebilir [2]. Bu algoritmaların arasında basit ve uygulanabilirliği kolay olması sebebiyle bölünmeli algoritmalar literatürde önemli yer tutmaktadır.…”

Section: Gi̇ri̇ş (Introduction)unclassified

“…Özellikle genetik algoritma [7], parçacık sürü optimizasyon [8], diferansiyel gelişim [9] ve yapay arı koloni algoritması [10] tabanlı geliştirilen literatürde birçok çalışma mevcuttur. Ancak geliştirilen bu yöntemlerin de yerel yakınsama problemlerini tamamıyla ortadan kaldırdığını söylemek mümkün değildir [2]. İlgili bölünmeli algoritmaların performansını arttırmaya yönelik öne sürülen bir diğer çözüm ise kernel fonksiyonları ile veri setlerinin linear uzaydan linear olmayan uzaya aktarılarak ele alınmasıdır.…”

Section: Gi̇ri̇ş (Introduction)unclassified

Diferansiyel Gelişim Tabanlı Çoklu Bulanık Kernel Kümeleme

Hançer

2018

Gazi Üniversitesi Mühendislik-Mimarlık Fakültesi Dergisi

Self Cite

View full text Add to dashboard Cite

Highlights:Graphical/Tabular Abstract  Using multiple kernel functions, a fuzzy clustering objective function is proposed.  The proposed objective function is integrated into the differential evolution framework.  A comparative analysis of the proposed clustering algorithm with well-known partitional clustering algorithms is performed. Figure A. The visual results of Haberman dataset.Purpose: The overall goal of this paper is to propose a multiple kernel fuzzy clustering approach using differential evolution (DE) to detect not only arbitrarily shaped clusters but also overlapping clusters in the data. Theory and Methods:The overall schedule of the proposed DE-based multiple kernel fuzzy clustering algorithm (DE-MKFC) is as follows. First, the data is first normalized and then transformed to the kernel space using a type of exponential function. Then, the population is randomly initialized with K number of cluster centroids. The quality of each individual is evaluated using the multiple kernel version of the fuzzy within-cluster distance. Results:To verify the effectiveness of the proposed DE-MKFC algorithm, it is analyzed and compared with representative clustering approaches, such as K-means, FCM, and MKFC on a variety of well-known benchmark datasets in terms of external validity indexes. The results show that DE-MKFC can detect non-arbitrary shaped clusters and generally achieves significantly better cluster performance than existing clustering algorithms. The effectiveness of DE-MKFC can also be illustrated in Fig. A. Conclusion:The main goal of this paper is to develop a multiple kernel fuzzy clustering algorithm. This aim is achieved by developing an objective function based on the principles of multiple kernel clustering and then integrating this objective function in DE framework. According to the results, the proposed algorithm can improve the quality of clusters and detect the overlapping clusters. It should be notified that multiple kernel fuzzy clustering is considered for the first time using an evolutionary technique.

show abstract

“…Therefore, in these application domains, techniques without requiring a user to input the number of clusters, and can also handle the different size and shape of genes. Meanwhile, density-based niche partitioning and sharing-based niche are used to maintain population diversity, capture the global optimal solution, and enhance exploitation capabilities [26,27], meanwhile, adaptive crossover and mutation probabilities are also employed to avoid the local optimum.…”

Section: Introductionmentioning

confidence: 99%

An Automatic K-Means Clustering Algorithm of GPS Data Combining a Novel Niche Genetic Algorithm with Noise and Density

Zhou

Shen

et al. 2017

IJGI

View full text Add to dashboard Cite

Rapidly growing Global Positioning System (GPS) data plays an important role in trajectory and their applications (e.g., GPS-enabled smart devices). In order to employ K-means to mine the better origins and destinations (OD) behind the GPS data and overcome its shortcomings including slowness of convergence, sensitivity to initial seeds selection, and getting stuck in a local optimum, this paper proposes and focuses on a novel niche genetic algorithm (NGA) with density and noise for K-means clustering (NoiseClust). In NoiseClust, an improved noise method and K-means++ are proposed to produce the initial population and capture higher quality seeds that can automatically determine the proper number of clusters, and also handle the different sizes and shapes of genes. A density-based method is presented to divide the number of niches, with its aim to maintain population diversity. Adaptive probabilities of crossover and mutation are also employed to prevent the convergence to a local optimum. Finally, the centers (the best chromosome) are obtained and then fed into the K-means as initial seeds to generate even higher quality clustering results by allowing the initial seeds to readjust as needed. Experimental results based on taxi GPS data sets demonstrate that NoiseClust has high performance and effectiveness, and easily mine the city's situations in four taxi GPS data sets.

show abstract

A comprehensive survey of traditional, merge-split and evolutionary approaches proposed for determination of cluster number

Cited by 108 publications

References 174 publications

MMKK++ algorithm for clustering heterogeneous images into an unknown number of clusters

MMKK++ algorithm for clustering heterogeneous images into an unknown number of clusters

Diferansiyel Gelişim Tabanlı Çoklu Bulanık Kernel Kümeleme

An Automatic K-Means Clustering Algorithm of GPS Data Combining a Novel Niche Genetic Algorithm with Noise and Density

Contact Info

Product

Resources

About