An efficient incremental clustering based improved K-Medoids for IoT multivariate data cluster analysis

Abstract

Clustering the data is an efficient way present in data analysis. Most of the clustering techniques are unable to underlying the hidden patterns, since the algorithms are supposed to store the data at a time from the data repository for data analysis. These data objects are infeasible, since the Internet of Things (IoT) dynamic data is too large to process and perform analysis over it. In olden days, the traditional clustering techniques implemented on batch processing systems with static data. In recent days, while considering IoT, Big data, and sensor technologies, the multivariate data is huge and unable to perform analysis with traditional approaches. Therefore, clustering multivariate data with an efficient way is a challenging problem and yielding insignificant clustering results. To overcome these limitations, in this paper, an Efficient Incremental Clustering by Fast Search driven Improved K-Medoids (EICFS-IKM) for IoT data integration and cluster analysis is proposed. The proposed EICFS-IKM contains cluster creating and cluster merging techniques for integrating the current dynamic multivariate data into the existing pattern data for final clustering data. For dynamically updating and modifying the centers of clusters of the new arriving instances, the improved k-medoids is employed. The proposed EICFS-IKM has implemented and experimented on four UCI machine learning data repository datasets, two dynamic industrial datasets, two linked stream datasets and compared with leading approaches namely IAPNA, IMMFC, ICFSKM, and E-ICFSMR and yielding encouraging results with computational time, NMI, purity and clustering accuracy.