Enhancing Community Detection for Big Sensor Data Clustering via Hyperbolic Network Embedding

Abstract

In this paper we present a novel big data clustering approach for measurements obtained from pervasive sensor networks. To address the potential very large scale of such datasets, we map the problem of data clustering to a community detection one. Datasets are cast in the form of graphs, representing the relations among individual observations and data clustering is mapped to node clustering (community detection) in the data graph. We propose a novel computational approach for enhancing the traditional Girvan-Newman (GN) community detection algorithm via hyperbolic network embedding. The data dependency graph is embedded in the hyperbolic space via Rigel embedding, making it possible to compute more efficiently the hyperbolic edge-betweenness centrality (HEBC) needed in the modified GN algorithm. This allows for more efficient clustering of the nodes of the data graph without significantly sacrificing accuracy. We demonstrate the efficacy of our approach with artificial network and data topologies, and real benchmark datasets. The proposed methodology can be used for efficient clustering of datasets obtained from massive pervasive smart city/building sensor networks, such as the FIESTA-IoT platform, and exploited in various applications such as lower-cost sensing.