A Novel clustering tendency assessment algorithm for WSN generated Spatio-Temporal data
Kartik Vishal Deshpande, Dheeraj Kumar · 2021
Wireless sensor networks are an important component of the internet of things implementation in several domains. These sensor networks generate large amount of spatiotemporal data, which need to be analyzed for knowledge creation. Exploratory analysis of this data using clustering (and assessing clustering tendency prior to actual clustering) can help understand underlying characteristics of the data. Most algorithms for clustering spatio-temporal data does not consider the spatial and temporal contiguity of datapoints before assigning them to clusters, leading to false results. In this paper, we propose nccVAT, a novel algorithm which assess clustering tendency of the spatio-temporal data and find contiguous data clusters for better understanding of the phenomenon being observed. The proposed algorithm is compared with state of the art algorithms such as ST-DBSCAN, ST-OPTICS, and clusiVAT, and Dunn's index is used to validate the results. The experiments performed on a real-life Intel Berkeley Research Laboratory (IBRL) dataset confirm superiority of nccVAT clusters as they have highest Dunn's index.