Synthetic Smartphone Data Usage Modeling Through User Clustering

Mobile network usage models (along with system models) are useful tools for estimating the performance of future and current mobile communication systems. However current mobile network usage models lack either the granularity or flexibility required for assessing advanced concepts (such as cognitive radio and dynamic spectrum access) at academic research levels. Therefore, in this work, we propose a network usage model that is both granular (modeling at the individual device level) and flexible (allowing arbitrary combinations of empirically derived user types and various time scales). We derive the user types through temporal clustering of network usage patterns of around 400 diverse US-based smartphone users. Additionally, the model accounts for the bursty nature and self-similarity of such traffic and we show these characteristics are preserved over several time scales.