For nearly 30-years, I have focused on the detection and reduction of financial fraud; specifically in payments. My research aims to reconcile symbolic AI and neural deep learning paradigms to move to the next generation of intelligent systems for fraud detection.
I am passionate about the practical application of machine learning and AI in business to really make a difference; teaching the next generation of "Data Whispers" is an honour.
I started my lifelong industry career in AI/Machine learning/Neural Computing in 1980s and was a pioneer of the practical deployment of machine learning in a diverse range of applications – with worldwide recognised “firsts”. I founded my own machine learning and data science company and grew this with offices in London, Singapore, Sydney and New York. I managed all aspects of the business with a focus on our research scientists and software development teams. Established myself as a leader in applying machine learning research to solve real-world problems.
I have a track record of computational neuroscience applied and deployed as solutions across industry, including credit/risk scoring (deployed in 35+ banks) and payment fraud detection (deployed at over 150+ institutions), which is today my main focus of research.
Following the sale of my business, I exited as the CEO in 2000 and continued my research, as well as a passion for teaching. Here, I work to deploy novel and practical AI, machine learning and deep learning to reduce financial crime with a focus on fraud detection in payment cards and AML.
Areas of specialism
University roles and responsibilities
- MSc Business Analytics (Machine Learning) - Teaching
- MSc Data Science (Practical Business Analytics) - Teaching
- Supervisor (MSc or PhD)
A Selection of Past AI/ML Industry Projects
- Identification of Customers Unlikely to Renew their Motor Breakdown Insurance using Neural Modelling– RAC
- Prediction of High-Spending Customers and their Product Types in Catalogue Company using Neural Data Mining– Freemans
- Blood Speckle Tracking using Neural Image Processing to Detect Cancer – Royal Marsden Hospital / Institute of Cancer Research
- Identification of Electricity Users to Predict Demand for Forward Purchasing of Electricity – Southern Electricity
- Oil Pipe Corrosion Detection – using non-intrusive sensors and neural signal processing – BP Oil (Alaska)
- Recognition of Coins for Vending Machines using Neural SOM Techniques with Low False-Positives – Mars Electronics
- Pizza Manufacturing Quality Control – Neural Image Processing – United Biscuits
- Quality Control in Manufacturing of Automotive Gear Box Assemblies using Laser Scanner and Neural Image Processing – Ford UK
- Identification of Defecting Photocopier Customers using Neural Data Mining – Rank Xerox
- Whisky Blending Analysis using a Novel bio-silicon Sensor and Neural DSP Processing – United Distillers
- Robot Control of Electric Wheelchair for Severely Disabled Persons Using Neural Computing and Fuzzy Logic for Navigation and User Input – TIC / DTI SMART Award
- Final Automatic Testing of DC Motors using Neural Digital Signal Processing – PMM plc.
- Locating Gold Ore Deposits in South Africa Using Neural Data Fusion and Anomaly Detection – Anglo America
- Early Detection of Wireless Telecom Subscribers Likely to Churn – Cellcom UK
- Early Prediction of Wireless Base-station Failure in Large UK Cellular Network using Neural Anomaly Detection – Orange plc.
- Optimising Delivery of Perishable Plants and flowers from Supplier/Depot to Shop – using Neural Combinatorial Optimisation (TSP[*]) Ring Network – Sainsbury’s Homebase
- Detection of Unusual Trading Behaviour and Fraud on the Trading Floor using Anomaly Detection – JP Morgan
- Prediction of FX US$ rates using Neural Networks – Citibank
- Scoring of Individuals for Automotive Loans to reduce Bad Debt – Lloyds Bowmaker
- Implementation of Hardware Parallel Processing System (16 Processors) for Real-Time Signal Analysis and recognition
- Implementation of a VMEbus[†]Based Image Processing Card with Neural Processor – Vero Microelectronics
- Speech Recognition system using HMM[‡]and Neural Networks to give Superior Speaker Independent Word Recognition in Noisy Environments
- SX-EW Process Optimisation for Copper Extraction
My research is focused on reducing and managing financial risk, fraud, anti-money laundering (AML) and all areas of cyber-crime. A neurocognitive approach with in context reasoning is proposed as a fundamental research area. I founded FITS as a not-for-profit “research institute” to tackle cyber-fraud, cyber-attacks and payment fraud:
- Using neural computing and adaptive techniques in the detection of emerging, real-time patterns in a high-volume transactional environment seeking to reconcile the symbolic and connectionist paradigms to move forward towards the next generation of intelligent systems.
- “Sea of Data” visualisation techniques to aid in the development of symbolic rules, models and “global” view.
- Automatic learning methodologies to create models for the detection of anomalous events and the automatic extraction of “rules” from models
- Inserting a neural network with fixed rules and parameters for known patterns and then automatically updating these during learning from on-line activities and presenting these changes to humans (emerging classes).
- Real-time deployment of complex neural hybrid systems in a mission-critical environment – including verification of models.
- Building timed events into a neural network architecture.
- Hybrid neural symbolic processing for the easy interpretation and knowledge extraction in real time, financial transaction environments, with robust fault-tolerant learning on in-complete or “noisy” data-feeds.
Postgraduate research supervision
MSc student dissertations in machine learning and data analytics. Please contact me if you are interested in working together.
Previous supervised student dissertations
- 2018, Andi Brajshori, Using Cost-Sensitive Artificial Neural Networks to Balance the Misclassification Costs for Fraud Detection through Threshold Moving
- 2018, Keerty Agarwal, Application of genetic algorithms for efficient field engineer scheduling: A case study for a leading energy and utilities provider in UK
- 2017, Giovanni Paolo Canuti, Deep Neural Network analysis of airlines taxi-out times at US airports.
Looking for students
I am looking to supervise PhD students; part time from industry very welcome.
Almost every business needs experts who can analyse their data making use of the latest machine learning and AI technologies. 80% of executives say this is creating jobs. Past students have joined companies such as Google, IBM, Tableau HSBC and British Gas based on these new skills. I will equip you with the skills necessary to kick-off an exciting career.
Only by combining data analytics with ML and AI can effective systems be created for problem-solving and making business decisions. In the US alone, there is a reported shortfall of 190,000 data analysts. In the UK, these roles are the highest job growth and attract some of the the highest salaries.
Courses I teach on
Cyber-fraud is lucrative to criminals, with a low risk of being convicted. There is a growing number of data breaches where criminals target legitimate computer systems to obtain large quantities of sensitive data that contain sufficient information so that they can then be used in payment card and other forms of bank fraud. My research over the last six years indicates that this situation will not change without changing the approach used to detect and manage fraud. My recent research focuses on the use of neural networks and AI to detect payment card fraud. I continue to work to bring a greater awareness of the issues and to promote strong research in collaboration with industry and the financial regulators. The level of fraud has been increasing since the introduction of payment cards and continues to increase. Our research has demonstrated that this is due to the nature of the complex data available, out-dated approaches to detection and management and the current economic and regulatory framework: providers see their losses through fraud as an acceptable cost to business. I believe that we can change that…