Big Network Analysis for Influence Identification on Social Networks

In the past few years there has been a dramatic increase in the generation of network-based data. Design and implementation of online marketing strategies based on this data could result in significant returns on investment. But the analysis of this data poses a significant challenge to managers due to: (1) the volume of data, (2) the signal-to-noise ratio in the data, and (3) limitations imposed by online social network providers. A common goal of marketers and brand managers with this data is the identification of influential individuals whom they wish to target with marketing campaigns. In this paper we study the suitability of big data technologies for network analysis. We break this analysis into two parts: (1) cloud-based solutions for analyzing large network datasets, and (2) optimal sampling methods for network analysis. We examine the suitability of distributed cloud computing resources for network analysis using different combination of hardware (GPU and memory optimized), and software (MapReduce and Spark). We also experiment with various network sampling techniques and examine the representativeness of commonly used sampling methods based on relevant network metrics