Stochastic Multi-Distribution Modeling of Inter-Contact Times

We present an accurate user-encounter trace generator based on an analytical model. Our method generates traces of intercontact times faster than models that explicitly generate mobility traces. In this encounter trace generation, we divide network agents in two groups based in their contact rates and assign them different probability distributions to better model pairs of agents with low contact rates. We added a periodical time-based term to the encounter trace generator to more accurately represent the preferential times for contact occurrence and assessed how it matches periodic patterns in the aggregate intercontact-time distribution and how it can recreate preferential routing times in the epidemic routing protocol. To validate this model we use previous results obtained in the literature and compare the generated encounter traces with the MIT reality mining experiment and obtained an average relative error of less than 1%. for both the aggregate intercontact-time distribution and for the gamma fit for the number of contacts per pair. Finally, we apply our trace-generation model to the epidemic routing protocol and provide both an analytical and simulation-based argument to show the computational efficiency of analytical encounter trace generators.