-the interface allows the expert system to tap data from the IMS cardholder databases hooked to an IBM mainframe
-when merchants call in for credit approval for a cardholder transmission, the expert system quickly pulls together and relevant database information, evaluates it, and makes either a decision or a recommendation as to whether or not credit should be granted.
-usually take into account as such factors as a cardholder's outstanding balance, payment history, buying habits, and the degree of risk involved
-should any warnings appear the system is even capable of asking merchants to obtain further information from the card holder such as a form on ID
-1/4 of all transactions processed by Authorizer's Assistant are straightforward enough to require no human intervention
-the system makes a decision and transmits it to the merchant
-the other transactions are referred to the system by human authorizers
-each is accompanied by an ID of the warning signs, an explanation regarding why the transaction is being referred, a recommendation for credit authorizer's or denial, decision making data culled from the data-bases.
-if the human authorizer disagrees with or questions a recommendation made by the expert system they can override it
-when humans and the expert system work together or decisions, the system cuts processing time by 20%
-to develop the knowledge base the developer uses a rule-based approach
-in doing so the developer's analysis interviews American Expresses 5 best authorizers in Ft. Lauderdale and created a knowledge base of 520 rules, the base was expanded to 800 after some refinement
-Authorizer's Assistant is still being used and supports more than 300 human authorizers
No comments:
Post a Comment