CMStatistics 2022: Start Registration
View Submission - CFE
A1561
Title: Identifying current account risk profiles to detect suspicious accounts Authors:  Rui Ying Goh - University of Edinburgh (United Kingdom) [presenting]
Galina Andreeva - University of Edinburgh (United Kingdom)
Yi Cao - University of Edinburgh (United Kingdom)
Johannes de Smedt - KU Leuven (Belgium)
Abstract: Anomalous financial behaviour signals a high exposure risk to suspicious activities, e.g. money laundering. Financial institutions build Know Your Client (KYC) profiles as the due diligence to first investigate typical account behaviour and then flag anomalous accounts which highly deviate from the typical ones. A two-stage approach is developed to assess anomaly risk from current account cash flow transactions. The first stage focuses on cluster analysis to explore normal current account profiles from the RFMP (Recency, Frequency, Monetary, Persistence) dimensions of transactional activities. The P dimension extends the popular RFM customer lifetime value marketing framework, to spot irregularities and detect anomalies. In the second stage, we evaluate the accounts with anomaly scores and examine the characteristics of high-risk accounts from two-dimensional visualisation plots and bag-of-word analysis of the transaction descriptions. The results reveal that the typical account profiles portray personal financial behaviour under different financial circumstances for day-to-day spending or saving purposes. We highlight potential suspicious account behaviour i.e., unusually large number of irregular or over-persistent transactions and disproportionate spending on transfer payments. These findings enhance KYC profiling, where financial institutions can spot accounts with high anomaly risk, before escalating the account owner's profile for further financial checks.