This session will explain how data science can be applied for detection of financial or other criminal and abusive activities. How do we catch terrorist financing? How to detect corruptioners? How bank can monitor your transactions? It will help the audience understands how data scientist approach fraud detection and what are the standard and advanced techniques and common pain points.
The first part of the talk will cover the analytical and mathematical background of the problematics. We will start with what are the data sources commonly used in the detection, describe the suitable supervised and unsupervised machine learning algorithms, say when they can be used and when not and end with how to approach optimization of existing fraud detection systems. Then we will explain how to apply this in the business context of detection of tax fraud, money laundering, money transfers and corruption. We also will cover the current status of the anti fraud industry in different countries and industries, which tools and methods companies
The second part of the talk will be practically oriented and explaining the application of data science in the past fraud detection projects from our experience. On examples from financial, public, healthcare or e-commerce we will uncover the different approaches to fraud detection. You will hear examples from simple rule base threshold models up to optimized multivariate detection.
This talk is suitable for both data scientist and business audience as it covers both analytical and business related aspects of the problematics. Finally, we will provide a workshop with a real life anonymous data to help you learn predicting a fraud cases, its key critical points and limitations.
Liubomyr is a Senior Consultant in Data Analytics practice of PwC. He participated in numerous projects working with Customer Analytics, Financial Crime Analytics, and Financial modeling. Prior to PwC, Liubomyr worked as data scientist on a project basis at banks, consulting companies, software vendors, and data science boutiques. He was responsible for statistical and econometric modeling and delivering the business insights. He studied Economics and Econometrics at the Center of Economic Research & Graduate Education (Charles University and Czech Academy of Science).
Richard is a data scientist with end-to-end experience with definition, design, PoC and implementation of various machine learning and risk management solutions and systems.
Last five years he was working in different roles from back-end developer to client facing consultant for various clients in financial industry worldwide.
Currently he is leading the PwC team developing a new risk scoring model for one of the largest Czech banks.
Event Timeslots (1)
Track B (Lower Floor)
Liubomyr Bregman, Richard Bobek