Neural Technologies is pleased to announce the launch of its EDR Classifier product, a tool which helps telecommunication operators to improve customer relations, whilst minimising revenue loss.
Although fraud in telecommunications networks represents a direct cause of revenue leakage, it also has a number of subtler indirect effects that cause network operators to incur additional losses. When a fraud is discovered, for example, it is usually not cost effective to identify which of the individual CDRs in the billing pipeline were part of the fraud because of the time and complexity of the analyses that would be required.
Since charging for fraud CDRs is unacceptable from a customer relations perspective, however, large numbers of CDRs corresponding to calls that were made around the time of a fraud are usually removed from the pipeline regardless of whether there is specific evidence to suggest that they were actually fraudulent or not. Since many of these CDRs will actually be chargeable, this process represents a source of revenue loss for the network operator.
To stop this loss, Neural Technologies has developed a neural network-based interactive decision support and visualisation tool that makes it possible to quickly and reliably identify individual fraud CDRs within an archive of thousands. Using a patented neural network engine, Nt's EDR classifier can process thousands of event data records per second on a desktop PC and performs smart context sensitive analyses to identify unusual activity.
The visualisation component of the EDR classifier presents the user with representations of a large number of characteristics of the CDRs within a CDR archive, such as whether they were associated with international calls, hot destination calls, sequential dialling, overlapping calls, unusually long or expensive calls, etc., and allows most fraud CDRs to be identified almost immediately by eye.
The EDR classifier is much more than a visualisation tool, however, because it makes intelligent context sensitive judgements about CDRs based on information provided by its user. Specifically, its user can provide it with hints by indicating CDRs that are not likely to be part of a fraud, and, optionally, CDRs that are. The EDR classifier neural network learns from this information and updates its classifications within seconds.
The speed, complexity, and richness of the analyses offered by the EDR classifier make it an enabling technology. For the first time the EDR classifier makes it possible for analysts to quickly and accurately identify individual fraud CDRs within a billing pipeline containing thousands of CDRs, therefore plugging an important hole in telecommunications revenue management strategies while simultaneously safeguarding customer relations.
For more information contact us.
| 
