Presented By O'Reilly and Cloudera
Make Data Work
December 1–3, 2015 • Singapore

Using machine learning to identify fraud on Telecom networks

Arshak Navruzyan (
4:50pm–5:30pm Wednesday, 12/02/2015
IoT and Real-time
Location: 324 Level: Intermediate
Tags: telecom
Average rating: ***..
(3.89, 9 ratings)

Prerequisite Knowledge

General familiarity of machine learning concepts and terms.


The Telecom industry faces a massive challenge: $46 billion is lost each year to sophisticated fraud attacks including call-back, international revenue sharing fraud, subscription fraud, etc.

We will discuss the use of unsupervised nonparametric methods for anomaly detection based on multivariate similarity. Such algorithms can be extremely effective in identifying changing fraud patterns (in the absence of labeled data),
as they are not trained on historical fraud cases but rather monitor “normal” subscriber behavior.

When labeled datasets are more commonly available, fraud departments often face a near-insurmountable imbalanced-class problem, as by definition fraud happens rarely and often goes undetected.

The talk will also address some recent supervised learning techniques that can deal with large imbalanced-class approaches for near-real-time identification of Telecom fraud.

Lastly, we will cover the production architecture of a machine learning system that can cope with the vast volumes of data flowing over carrier networks, to provide near-real-time prediction and minimize losses due to fraud.

Photo of Arshak Navruzyan

Arshak Navruzyan

Arshak is a machine learning focused product manager. He founded Fellowship.AI applied machine learning fellowship program and is a cofounder of Platform.AI.

He has delivered AI solutions for some of the largest enterprises in the world and multi-billion dollar quantitative hedge funds.

Previously Arshak served as the Chief Technology Officer at Sentient Technologies. He has also been in technology leadership roles at Argyle Data, Alpine, Endeca/Oracle.

Comments on this page are now closed.


Picture of Sonal Goyal
Sonal Goyal
12/16/2015 5:30am +08

Are the slides for this talk available?