Periodicity Detection of Node Behaviour in Opportunistic Mobile Social Networks
Bambang Soelistijanto, Elisabeth Permatasari · 2019
The recent rise of networks that rely on human mobility, such as opportunistic mobile social networks (OMSNs), has prompted the need for methods that detect the periodic patterns of node movements. Knowledge of the periodicity of node behaviour is essential to design effective and efficient network protocols in such networks. Node behaviour in OMSNs is typically characterized by the node contact patterns. In fact, node connections in these networks occur intermittently, resulting in sparse contact data. Consequently, the traditional periodicity detection methods, e.g. the FFT periodogram and autocorrelation, that favour complete, regularly-sampled time-series data are unsuitable in this setting. In this paper, we exploit the Lomb-Scargle periodogram, initially designed to handle incomplete or irregular sampling data, to identify node behaviour periodicity in OMSNs. Using simulation driven by real human contact traces, we show that the technique is able to accurately detect the behaviour periodicity of majority nodes in the network, even for those with a high level of sparsity contact data.