Aim

Using Bayesian graphical modelling for combining evidence to assess the likelihood of real-world events.

Objectives

1. Develop graphical modelling approaches for linking together explanations and evidence for sensing applications.
2. Develop Bayesian statistical methodology for combining together evidence from different sources, including both subjective and quantitative evidence.
3. Apply this methodology to real sensing applications to evaluate its performance.

Description

In sensing applications, it is frequently necessary to bring together measurements (which may be uncertain) with other sources of evidence such as computer models of the relevant system, and background intelligence to try and understand what is happening on the ground. This project will develop an approach based on chain event graphs for combining evidence of different natures from different sources with possible explanations for this evidence. These graphs give a clear visual picture of the ways in which the explanations and the evidence interact through time so that non-statisticians can see the evolution of the event. This approach will draw on previous work in forensic statistics, where chain event graphs have been used to establish the likelihood of different hypotheses given different sources of evidence in criminal cases.

References

Gail Robertson, Amy L Wilson, Jim Q Smith, Chain event graphs for assessing activity-level propositions in forensic science in relation to drug traces on banknotes, Law, Probability and Risk, Volume 23, Issue 1, 2024, mgae013, https://doi.org/10.1093/lpr/mgae013