Cross-Domain Signal Extraction Using Sparse Network Sampling

This research project seeks to adapt signal processing algorithms developed for the detection and estimation of weak RF network signals to the application of detecting weak social media signals diffusing across multiple, heterogeneous social networking environments (SNEs). We propose to develop quantitative predictive models that capture the dynamics of information diffusion processes over multiple SNE domains to include the cross-domain ripple effect, where a social trending topic in one domain can penetrate into another domain. We will leverage the predictive models to accurately detect early in the diffusion process weak social media signals that will propagate widely in the future. The IvySys model-based predictive analytics approach exploits the fact that orchestrated and automated efforts to spread information across SNEs can produce different patterns. The IvySys approach uses signal-processing techniques, such as Kalman filtering, to adaptively estimate cross-domain model parameters, representing the underlying dynamics of the social diffusion process, in real time using sampled streams of SNE data.