An efficient fuzzy c-means approach based on canonical polyadic decomposition for clustering big data in IoT

Mining smart data from the collected big data in Internet of Things which attempts to better human life by integrating physical devices into the information space. As one of the most important clustering techniques for drilling smart data, the fuzzy c-means algorithm (FCM) assigns each object to multiple groups by calculating a membership matrix. However, each big data object has a large number of attributes, posing an remarkable challenge on FCM for loT big data real-time clustering. In this paper, we propose an efficient fuzzy c-means approach based on the tensor canonical polyadic decomposition for clustering big data in Internet of Things. In the presented scheme, the traditional fuzzy c-means algorithm is converted to the high-order tensor fuzzy c-means algorithm (HOFCM) via a bijection function. Furthermore, the tensor canonical polyadic decomposition is utilized to reduce the attributes of every objects for enhancing the clustering efficiency. Finally, the extensive experiments are conducted to compare the developed scheme with the traditional fuzzy c-means algorithm on two large loT datasets including sWSN and eGSAD regarding clustering accuracy and clustering efficiency. The results argue that the developed scheme achieves a significantly higher clustering efficiency with a slight clustering accuracy drop compared with the traditional algorithm, indicating the potential of the developed scheme for drilling smart data from loT big data. (C) 2018 Elsevier B.V. All rights reserved.