Public Funding for Performance Horizon Group Limited

To embed sophisticated intelligent systems methods leading to a commercialisation of a new fraud detection product.

The TRINITY project will undertake research into the analytics algorithms required for realtime data mining of large volumes of retail-based online purchasing transactions. The objectives are to create/adopt algorithms to quickly: a) provide anomaly/fraud detection; b) detect invalid integration in implementations of e-commerce web sites; c) enable process optimisation by analysing a broader set of event oriented information; and d) identify the technical issues to be resolved for providing a significantly more refined payments tracking mechanism where money has to flow back through a complex chain of suppliers. Within six months these new algorithms must analyse a minimum of 3-5 million transactions per day (current state-of-the-art processing typically handles a total of 1.5 million sales per month) i.e. a factor of 60-90 increase. A further 10 fold increase is expected six months later. The associated reports must be produced within seconds (some reports are used to decide payment to various parts of the supply chain hence the need for accuracy and timeliness). Successful research into, and deployment of, these new analytics will enable online vendors to more effectively: a) track payment expectations through more accurate understanding of how a sale is achieved for even the most complex combinations of channels to market and through a number of valueadd intermediaries; b) detect and avoid new fraudulent payment schemes in which professional criminals use a combination of co-ordinated and opportunistic mechanisms to obtain false payments; and c) create, manage and improve their e-commerce web-sites by using the reports that identify any functional flaws in the web-sites and which provide insight into ways in which the associated business process can be improved. The key deliverable of this six-month research project is the evaluation of the new set of data algorithms through their deployment and evaluation in a Proof of Concept demonstrator.