Artificial intelligence technology (Machine Learning) and mathematical data mining algorithms contained in PSFraud, as well as the use of the Symbolic Vector to monitor the atypicality of user behavior, allow the identification of fraud with the greatest certainty. In addition, the compact symbolic vector storage scheme and the parallelism of the systems allow to extend the processing to loads of any size, being able to analyze millions of transactions in a few seconds, i.e., practically in real time.
In order to obtain these results, PSFraud incorporates the best-known computing techniques, combining them in a unique way, producing better detection rates with lower alert levels.
The ability to learn continuously from incoming transactions allows the user profile of the Bantotal banking system to be continuously updated, in order to assimilate knowledge of behavioral patterns with their evolution and automatically update the rules and parameters necessary to stop fraud. The behavior pattern of the banking system user is calculated continuously, always looking for strong changes that indicate atypical behavior which, ultimately, may be associated with a fraud attempt by a third party.
PSFraud’s self-learning capability significantly reduces false alarms, making it less necessary for human effort to monitor them, as well as reducing associated indirect costs.