Managed real-time transaction fraud analysis and decisioning

A method of processing a payment transaction, comprising:

• generating a rule for use in determining if authorization should be declined for the payment transaction, wherein the process of generating the rule further comprises:
• selecting a plurality of payment transaction risk assessment models, each risk assessment model producing a score indicative of the likelihood of a transaction being fraudulent;
• for each risk assessment model, dividing a range of the score into multiple intervals;
• and
• generating the rule by combining an interval of the range of a first risk assessment model with an interval of the range of a second risk assessment model;
• and applying the rule to the payment transaction to determine if authorization for the transaction should be declined.

A method of generating a rule base for determining whether to authorize a payment transaction, comprising:

• selecting a plurality of payment transaction risk assessment scoring models, wherein each scoring model produces a value indicative of the risk associated with a transaction, the value falling within a range of values;
• for each scoring model, dividing the range of values into a plurality of intervals, wherein each interval represents a sub-range of the range of values;
• generating a proposed rule by selecting a combination of an interval of the range of a first scoring model and an interval of the range of a second scoring model;
• evaluating the performance of the proposed rule by applying the rule to data for previously completed transactions;
• and
• including the rule in the rule base if the performance of the rule meets or exceeds a specified criteria.

An apparatus for identifying payment transactions that should considered as potentially fraudulent, comprising:

• a processor configured to execute a set of instructions;
• a memory coupled to the processor for storing the set of instructions;
• and
• the set of instructions stored in the memory, wherein when executed by the processor the instructions implement a process to
• generate a rule for identifying potentially fraudulent payment transactions, wherein generating the rule further comprises
• selecting a plurality of payment transaction risk assessment models, each risk assessment model producing a score indicative of the likelihood of a transaction being fraudulent;
• for each risk assessment model, dividing a range of the score into multiple intervals;
• and
• generating the rule by combining an interval of the range of a first risk assessment model with an interval of the range of a second risk assessment model;
• access transaction data for a plurality of payment transactions;
• apply the generated rule to the transaction data for each of the plurality of payment transactions;
• identify a payment transaction that is one of the plurality of payment transactions as potentially fraudulent if the transaction data for that payment transaction satisfies the generated rule;
• and
• create a fraud case file for each payment transaction identified as being potentially fraudulent.