Spark-Based Measurement and Analysis on Offline Mobile Application Market over Device-to-Device Sharing in Mobile Social Networks

Recently how to select seeding users with large impacts in social networks has gained more and more attention in many studies. It has been demonstrated in a number of researches that the seeding users, playing vital roles in social groups, could be exploited to promote the dissemination of popular contents. Nevertheless, the existed algorithms, which are performed on small-scale data sets with limited feature dimensions, mostly base on unconsolidated hypotheses and measurements of data sets. Consequently, their performance and precision cannot be well evaluated and improved. In this paper, we firstly make comprehensive and large-scale measurements on 3.56 TBytes of real data sets related to Device-to-Device (D2D) content sharing activity traces from a popular D2D sharing application (APP). The mobile social networks generated by the offline content deliveries between users are presented and analyzed. Focusing on the seeding users' selection problem in social networks, we propose algorithms of weighted SeedRanks (SRs) to select the seeding users with accuracy. The algorithms are adapted to the parallel computing platform of Apache Spark with high performance. The results of the recurrent experiment on the large-scale D2D data sets prove the efficiency of our algorithms. Finally, we make conclusions and discuss future work.