Extending fuzzy c-means to clustering data streams

Abstract

A data stream is an ordered and continuous sequence of examples that can be examined only once. Data stream mining introduces new challenges compared to traditional mining algorithms. Fuzzy c-means (FCM) is a method of clustering in which a data point can assign to more than one cluster at the same time. In this paper we extend FCM algorithm to clustering data streams. Our performance experiments over KDD-CUP'99 data set show the efficiency of the algorithm.

Merging Clusters in Summary Structures for Data Stream Mining based on Fuzzy Similarity Measures

“…This algorithm divides the data set into chunks and clusters each chunk in sequence using the Weighted Fuzzy C-Means algorithm (WFCM) [4]. The weighted FCM -Adaptive Cluster [15] and Online Fuzzy C-Means [11] are examples of algorithms based on this approach. A survey on fuzzy methods for data streams clustering can be found in [1].…”

Clustering of nonstationary data streams: A survey of fuzzy partitional methods

“…The problems of detecting outliers and tracking concept change are not considered in the previously described methods, where concept change is handled by reclustering. To cope with evolving nature of data streams, an extension to sWFCM called WFCM with adaptive cluster number (wFCM‐AC; Mostafavi & Amiri, ) was presented. The adaptive process is a local search by varying the number of clusters, as summarized in algorithm 1.…”

“…Fuzzy clustering methods of data streams Probabilistic (mean) Probabilistic (medoid) Possibilistic SFCM (Hore et al, 2007a) SPFCM (Hore et al, 2007b) OFCM (Hore et al, 2008) sWFCM (Wan et al, 2008) wFCM-AC (Mostafavi & Amiri, 2012) HOFCMD (Labroche, 2014) OFCMD (Labroche, 2014) IMMFC (Wang et al, 2014) wPCM stream (Jaworski et al, 2012) eGKPCMPCM (Škrjanc & Dovžan, 2015) TRAC-STREAMS (Nasraoui & Rojas, 2006) GPCM stream FIGURE 2 Fuzzy clustering methods of data streams…”

Performance Comparison of Two Streaming Data Clustering Algorithms

“…Mostafvi and Amiri [29] extended WFCM algorithm and called it. In WFCM-AC initially FCM is applied to the normalized chunk of stream data, that each data point is replaced by subtracting the total mean and dividing the result by the standard deviation.…”

“…In WFCM-AC initially FCM is applied to the normalized chunk of stream data, that each data point is replaced by subtracting the total mean and dividing the result by the standard deviation. Membership value of each point is represented by membership matrix [29]. Each obtained center can be weighted by summing the membership values of all examples that have partial belonging to it.…”