Redescription, Malleable Visualization, and Storytelling for Causal Data Mining

Harmonia proposes creating a dynamic Bayesian networks and redescription/story telling (DBN-RS) tool to help solve problems such as the problem of condition-based maintenance for vehicles. DBN-RS offers two means of data mining. The first is DBN which is used in an automated mode on discrete event streams, and by applying an algorithm that marries frequent pattern mining with probabilistic modeling (DBN). DBN includes frequent pattern mining which is scalable to large data sets but does not exhibit the rigorous probabilistic interpretations that are the mainstay of the graphical models literature. DBN also includes probabilistic modeling providing a formal probabilistic basis to model relationships between time-indexed random variables but is intractable to learn in the general case – and hence they do not normally scale well. The second means of data mining is an interactive visualization method that permits human guided exploration of data to try to “connect the dots” and see if there is a relationship between entities or events (RS). The goal of this research is to identify causal relationships from large data sets using these means of data mining in one tool.