Accelerating Fuzzy-C Means Using an Estimated Subsample Size

Abstract

Many algorithms designed to accelerate the Fuzzy c-Means (FCM) clustering algorithm randomly sample the data. Typically, no statistical method is used to estimate the subsample size, despite the impact subsample sizes have on speed and quality. This paper introduces two new accelerated algorithms, GOFCM and MSERFCM, that use a statistical method to estimate the subsample size. GOFCM, a variant of SPFCM, also leverages progressive sampling. MSERFCM, a variant of rseFCM, gains a speedup from improved initialization. A general, novel stopping criterion for accelerated clustering is introduced. The new algorithms are compared to FCM and four accelerated variants of FCM. GOFCM's speedup was 4-47 times that of FCM and faster than SPFCM on each of the six datasets used in experiments. For five of the datasets, partitions were within 1% of those of FCM. MSERFCM's speedup was 5-26 times that of FCM and produced partitions within 3% of those of FCM on all datasets. A unique dataset, consisting of plankton images, exposed the strengths and weaknesses of many of the algorithms tested. It is shown that the new stopping criterion is effective in speeding up algorithms such as SPFCM and the final partitions are very close to those of FCM.

Performance enhancement of swarm intelligence techniques in dementia classification using dragonfly‐based hybrid algorithms

“…[29][30][31][32] The advantages of FCM over k-means or any other clustering algorithm is that an object can lie under any number of clusters with or without overlap and its intent of belongingness can be estimated according to the following constraints with n, and c being number of data samples and clusters, respectively: [29][30][31][32] The advantages of FCM over k-means or any other clustering algorithm is that an object can lie under any number of clusters with or without overlap and its intent of belongingness can be estimated according to the following constraints with n, and c being number of data samples and clusters, respectively:…”

“…FCM is one among the popular clustering algorithm, which samples the data without following the statistical approach for determining the size of the subsample. [29][30][31][32] The advantages of FCM over k-means or any other clustering algorithm is that an object can lie under any number of clusters with or without overlap and its intent of belongingness can be estimated according to the following constraints with n, and c being number of data samples and clusters, respectively:…”

BigFCM: Fast, precise and scalable FCM on hadoop

“…Since FCM choose the initial centers randomly, the final result and especially its convergence speed significantly depends on the original center selection. A method proposed to address this problem is based on estimated subsample size to improve the initialization [18]. In the field of clustering large amounts of data, three types of methods have been proposed:…”

Performance Analysis of Improved Clustering Algorithm on Real and Synthetic Data

“…First is the distance measure strategy and second is initial centroids selection strategy to minimize processing speed and increase stability. Paper [4] introduces two accelerated clustering algorithms using estimated subsample size and the novel stopping criterion. Authors in the paper [5] present a systematic study of kmeans-based consensus clustering algorithm, identify necessary and sufficient conditions for the algorithms on both pure and noisy datasets.…”

“…The authors in the paper [12] propose two novel enhanced algorithms such as geometric progressive fuzzy c-means and minimum sample estimate random fuzzy c-means by using some statistical techniques. This is to compute the size subsamples.…”