cyberquantic logo header
EN-language img
FR-language img
breadcrumbs icon
Banking & Finances

Fraud Detection

Detection of frauds based on collusions

Detection of frauds based on collusions

For:
Stock market regulator, stock traders, stock investors
Goal:
Automate a Business Process
Problem addressed
Automatic unsupervised detection of frauds based on collusions
Scope of use case
Validating the predicted collusion set is effort-intensive and investigative and
legal expertise are necessary
Description
Frauds are prevalent across all industries; and they are
particularly severe in todays computerized, web-connected,
mobile-accessible, and cloud-enabled business
environments. A Federal Bureau of Investigation (FBI)
report states that the insurance industry in the US, which
consists of over seven thousand companies and collects over
one trillion dollars in premiums, loses about forty billion
dollars annually in frauds in the non-health insurance sector
alone. The aggregate size of the 52 regulated stock
exchanges across the world (total market capitalization) was
$55 trillion as of December 2012. Given the money involved,
it is not surprising that the stock market is a target of frauds.
Many malpractices in stock market trading, e.g. circular
trading and price manipulation, use the modus operandi of
collusion. Informally, a set of traders is a candidate collusion
set when they have heavy trading among themselves, as
compared to their trading with others. We formalize the
problem of detection of collusion sets, if any, in a given
trading database. We show that na�ve approaches are
inefficient for real-life situations. We adapt and apply two
well-known graph clustering algorithms for this problem.
We also propose a new graph clustering algorithm,
specifically tailored for detecting collusion sets; further, we
establish a combined collusion set. Treating individual
experiments as evidence, this approach allows us to quantify
the confidence (or belief) in the candidate collusion sets. We
have carried out detailed simulation experiments to
demonstrate effectiveness of the proposed algorithms. The
system is also operational in a government organization.
Note that all our collusion detection algorithms are
completely unsupervised and do not need any training data.
understand frame img
Machine Learning
Interested in the same or similar project?
Submit a request and get a free project evaluation.