I am Professor in Computer Science and Statistics in Queen Mary, University of London. My research interests cover Bayesian modeling and risk quantification in diverse areas. My experience in applying Bayesian methods to real problems has convinced me that intelligent risk assessment and decision analysis requires knowledge and data. Not just "Big Data". I am also a joint founder and of Agena Ltd, who develop and distribute AgenaRisk, a software product for modeling risk and uncertainty. At Queen Mary I teach decision and risk analysis and software engineering.
I was a fellow at the Newton Institute for Mathematical Sciences, Cambridge University in 2016 and was invited to the Fields Institute for Research in Mathematical Sciences, University of Toronto, Canada in 2010. I was also a fellow at the Isaac Newton Institute for Mathematical Sciences, Cambridge University in 2016 on Probability and Statistics in Forensic Science.
I am interested in intelligent risk assessment and decision analysis using knowledge and data. Typically this involves analysing and predicting the probabilities of unknown events using Bayesian statistical methods including causal, probabilistic models (Bayesian networks). In addition to working on theoretical and algorithmic foundations, this work covers a wide range of application domains such as medical analytics, legal reasoning, embedded software, operational risk in finance, systems and design reliability (including software), project risk,commercial risk, decision support, cost benefit analysis, AI and personalization, machine learning, legal argumentation, cyber security and football prediction.
My involvement in funded research projects, as principal or co-investigator includes:
Most of my technology transfer and applications activity is pursued through Agena Ltd. For the last 15 years I have been instrumental in developing the AgenaRisk software platform, written in Java and available as a sophisticated desktop tool, for Windows, Macintosh and Linux operating systems. An API version is available too. A free version of AgenaRisk is available to download and a Lite version is also available if you purchase the book - this has extended functionality. You can find out more about these activities here:
I have consulted or given training to many organisations including General Electric, Milliman, Mastercard, AIB bank, Philips, NATS, QinetiQ, Advantica, Motorola, DSTL (UK MOD), ABSA, AON, Ericsson, Royal Bank of Canada, TNO (the Netherlands), BHP Billiton and ICRAF (The World Agroforestry Centre) and more.
This screen shot shows a Bayesian Network running in AgenaRisk. The nodes represent uncertain variables with probability distributions. The arcs represent functional or causal interactions. Building a model like this takes minutes compared to attempting it in MatLab or any other environment, and AgenaRisk is the only Bayesian Network application that can handle continuous variables properly, meaning you can build hybrid Bayesian networks. The great thing about this approach is that we can combine machine/statistical learning with subject matter expertise about the way the world works.
My book Risk Assessment and Decision Analysis with Bayesian Networks , co-authored with Norman Fenton, covers the majority of my work on Bayesian networks theory and applications. It was picked up some rather good reviews:
… although there have been several excellent books dedicated to Bayesian networks and related methods, these books tend to be aimed at readers who already have a high level of mathematical sophistication … . As such they are not accessible to readers who are not already proficient in those subjects. This book is an exciting development because it addresses this problem. … it should be understandable by any numerate reader interested in risk assessment and decision making. The book provides sufficient motivation and examples (as well as the mathematics and probability where needed from scratch) to enable readers to understand the core principles and power of Bayesian networks. However, the focus is on ensuring that readers can build practical Bayesian network models … readers are provided with a tool that performs the propagation, so they will be able to build their own models to solve real-world risk assessment problems.
So, I must admit. This book is nothing like what I expected when I ordered it. Having a background in engineering, many statistics classes (some graduate level), and experience with other types of "networks," I was expecting a fairly tough book filled with theorems, equations, and trivial first-principle examples. Was I ever wrong. Instead, this book discusses Bruce Willis blowing up a meteor, game shows, the author sleeping too long and missing work, and scenarios in Agatha Christie novels. In my stuffy statistics classes, Bayesian statistics was explained as something not short of gnosticism - that if you didn't get it already, it was not worth explaining. I used Bayes' Theorem for simple problems, but for nothing very practical - practical implementation was too difficult. It is a sign that someone knows a complex subject very well: Explain it simply enough (and still be correct) so that a bright grade-school child would understand. This is exactly that the authors have done. The appealing and simple approach to this book hides the fact that they to an amazingly good job at explaining Bayesian statistics. The light subjects I mentioned in the paragraph above should be taken that these authors do a tremendous job at explaining themselves, not that this book is a pushover, or is simple or pedestrian. If you, like me, have always wished that someone would have taken the time to explain Bayesian statistics to you, or if you are somewhat experienced in statistics but feel like you need more (or see the holes that these authors point out), then this book is for you. If you are looking for a book filled with proofs and theorems, or if you like authors that make themselves sound smart by intimidation and shock-and-awe, then skip this book. This book is the real deal.
My paper with Norman Fenton "A critique of software defect prediction models" placed in top 1% most influential papers in its field based on number of citations (according to Essential Science Indicators).