System and method for fraud detection
AssigneeACI Worldwide Corp.
InventorVladimir Ivakhnenko
Life of this patent
Estimating the term…
Abstract
A system and method for detecting fraudulent transactions of incremental transactions via incrementally training classifiers on a number of data subsets to obtain transaction label predictions. Recent information on fraudulent transactions is implemented to update the classifiers, thereby eliminating the problems of concept drift and verification latency. Some of the incremental transactions may be labeled as fraudulent according to the incrementally modified classifiers and thereby rejected.
Claims13
A system for detecting a fraudulent transaction, the system comprising:
- an electronic memory element containing a database storing an initial dataset corresponding to a plurality of initial transactions and a plurality of incremental datasets each corresponding to a plurality of incremental transactions;
- and
- a processor configured to:
- incrementally apply at least two classifiers based on the initial datasets to the incremental transactions of the incremental datasets, each classifier corresponding to a different learning algorithm;
- incrementally build at least two subsets per incremental dataset, each subset including a plurality of labels according to application of one of the at least two classifiers;
- incrementally modify the at least two classifiers before applying the at least two classifiers to incremental transactions of subsequent incremental datasets by applying one of the different learning algorithms not corresponding to the classifier being modified to the at least two subsets such that each classifier is updated via a learning algorithm corresponding to another one of the at least two classifiers;
- apply a label indicating fraudulence to one of the incremental transactions via the incrementally modified classifiers and the learning algorithms;
- and
- reject the fraudulent-labeled incremental transaction.
A method of detecting fraudulent transactions, the method comprising the steps of:
- receiving an initial dataset corresponding to a plurality of initial transactions;
- receiving a plurality of incremental datasets each corresponding to a plurality of incremental transactions;
- applying at least two classifiers based on the initial datasets to the incremental transactions of the incremental datasets, each classifier corresponding to a different learning algorithm;
- building at least two subsets per incremental dataset, each subset including a plurality of labels according to application of one of the at least two classifiers;
- incrementally modifying the at least two classifiers before applying the at least two classifiers to incremental transactions of subsequent incremental datasets by applying one of the different learning algorithms not corresponding to the classifier being modified to the at least two subsets such that each classifier is updated via a learning algorithm corresponding to another one of the at least two classifiers;
- applying a label indicating fraudulence to one of the incremental transactions via the incrementally modified classifiers and the learning algorithms;
- and
- rejecting the fraudulent-labeled incremental transaction.
Dependencies are inferred from the claim text.