Speaker: Miroslav Cepek - ORACLE LABS

Abstract: There are many problems where data can be naturally represented as a graph — network traffic analysis, financial transaction monitoring, social relationships, to name a few. In this talk, I will discuss applying a graph-based approach to support financial crime detection in banks. In a bank, there is an automated, rule-based mechanism to identify potentially problematic behaviour. The identified behaviour is called a case. The case is then investigated by a human analyst, who decide whether the case is benign, or the case should be reported to the proper authorities. We support analysts by identifying past cases similar in structure, to the present one. This allows the analyst to get insights from the past cases to get to correct decision faster. The talk will be divided into three parts - in the first I will talk about "Parallel Graph AnalytiX" — a graph database developed by Oracle Labs team, used to store cases. In the second part, I will talk about machine learning techniques for node and graph vector representation, and in the last, I will show how we applied mentioned methods to the problem of case similarity.

Short bio:Miroslav Cepek PhD, is presently working as Data Scientist in Oracle Labs on PGX project. His main focus is applying graph oriented machine learning algorithms to problem in Financial Crime and Compliance area. Most importantly building a solution to help to detect cases of money laundering in large banks. Before Oracle Labs Miroslav worked in Vendavo, a leading B2B pricing solution provider, working on intelligent margin and revenue analyses. Helping Fortune 2000 companies to sell profitably and set fair prices in global B2B environment. Miroslav gained his PhD in computer science in Faculty of Electrical Engineering in Czech Technical University and during his studies he also worked there as assistant professor working in Machine Learning field.

Download the poster.