Combining Existential Rules with Network Diffusion Processes for Automated Hypothesis Generation

There are many applications in which decision support systems can benefit from the ability to automatically generate hypotheses -- two salient examples are detection of malicious behavior (such as in social media platforms or cyber threat analysis in enterprise systems) and preventive healthcare that continuously monitors patients’ readings towards early detection of conditions requiring medical attention. In this paper, we continue work on the recently-proposed NETDER architecture (Network Diffusion and ontological reasoning based on Existential Rules), addressing the technical issues towards its effective implementation. The working hypothesis behind our model is that three key elements can be leveraged towards the automatic generation of hypotheses: (i) combining multiple, ever-evolving data sources, (ii) maintaining a knowledge base using logic-based formalisms capable of value invention to support reasoning about unknown objects based on available data, and (iii) maintaining a related knowledge base for actors and relationships among them, and how information flows across such a network. After developing the formal machinery behind the model, we focus on the main task of query answering (QA), studying three different policies that can be used towards this end. One of our main contributions in this regard is the analysis about computational cost of the proposed QA procedures, including cases in which termination of associated procedures is not guaranteed. Finally, we present a use case showing how the formalism can be applied in a real-world setting involving applications to cybersecurity.