Method and system for performing fraud detection for users with infrequent activity
First Claim
1. A non-transitory computer readable medium having instructions stored thereon which, when executed by a computer processor, cause the computer processor to perform operations comprising:
- for each previous transaction among one or more historical transactions for a party;
determining, by the computer processor configured to automate categorization, a similarity value between a current transaction for the party and the previous transaction, wherein the similarity value is determined by computing an initial weight for each of the properties of a set of properties, computing a similarity between each of the properties of the current transaction and the properties of the previous transaction, adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity value by summing the products of the normalized adjusted weights and the computed similarities;
determining, by the computer processor configured to automate categorization, that the similarity value is greater than or equal to a predetermined threshold value;
categorizing, by an anomaly confidence generator component of the computer processor, the current transaction as not anomalous in response to determining that the similarity value is greater than or equal to the predetermined threshold value;
determining, by the computer processor configured to automate categorization, a factor for the previous transaction based on an age of the previous transaction;
computing, by the computer processor configured to automate categorization, a rank for the previous transaction using the similarity value and the factor; and
storing, by the computer processor configured to automate categorization, the computed rank in a database configured to store the computed rank; and
computing, by the computer processor configured to automate categorization, a confidence in the categorization of the current transaction as not anomalous based on the stored ranks, wherein the confidence indicates whether the current transaction for the party is a fraudulent transaction, and wherein the current transaction is an Internet log-in.
7 Assignments
0 Petitions
Accused Products
Abstract
A method of categorizing a recent transaction as anomalous includes a) receiving information about a recent transaction and b) accessing information about one or more historical transactions. The one or more historical transactions have at least one party in common with the recent transaction. The method also includes c) determining a similarity value between the recent transaction and a transaction i of the one or more historical transactions and d) determining if the similarity value is greater than or equal to a predetermined threshold value. The method further includes e) if the similarity is greater than or equal to the predetermined threshold value, categorizing the recent transaction as not anomalous or f) if the similarity is less than the predetermined threshold value, determining if there are additional transactions. If there are additional transactions, incrementing counter i and repeating steps c) through f).
16 Citations
28 Claims
-
1. A non-transitory computer readable medium having instructions stored thereon which, when executed by a computer processor, cause the computer processor to perform operations comprising:
-
for each previous transaction among one or more historical transactions for a party; determining, by the computer processor configured to automate categorization, a similarity value between a current transaction for the party and the previous transaction, wherein the similarity value is determined by computing an initial weight for each of the properties of a set of properties, computing a similarity between each of the properties of the current transaction and the properties of the previous transaction, adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity value by summing the products of the normalized adjusted weights and the computed similarities; determining, by the computer processor configured to automate categorization, that the similarity value is greater than or equal to a predetermined threshold value; categorizing, by an anomaly confidence generator component of the computer processor, the current transaction as not anomalous in response to determining that the similarity value is greater than or equal to the predetermined threshold value; determining, by the computer processor configured to automate categorization, a factor for the previous transaction based on an age of the previous transaction; computing, by the computer processor configured to automate categorization, a rank for the previous transaction using the similarity value and the factor; and storing, by the computer processor configured to automate categorization, the computed rank in a database configured to store the computed rank; and computing, by the computer processor configured to automate categorization, a confidence in the categorization of the current transaction as not anomalous based on the stored ranks, wherein the confidence indicates whether the current transaction for the party is a fraudulent transaction, and wherein the current transaction is an Internet log-in. - View Dependent Claims (2, 3, 4, 5, 6)
-
-
7. A computer-implemented method comprising:
-
for each previous transaction among one or more historical transactions for a party; determining, by a computer processor configured to automate categorization, a similarity value between a current transaction for the party and the previous transaction, wherein the similarity value is determined by computing an initial weight for each of the properties of a set of properties, computing a similarity between each of the properties of the current transaction and the properties of the previous transaction, adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity value by summing the products of the normalized adjusted weights and the computed similarities; determining, by the computer processor configured to automate categorization, that the similarity value is greater than or equal to a predetermined threshold value; categorizing, by an anomaly confidence generator component of the computer processor, the current transaction as not anomalous in response to determining that the similarity value is greater than or equal to the predetermined threshold value; determining, by the computer processor configured to automate categorization, a factor for the previous transaction based on an age of the previous transaction; computing, by the computer processor configured to automate categorization, a rank for the previous transaction using the similarity value and the factor; and storing, by the computer processor configured to automate categorization, the computed rank in a database configured to store the computed rank; and computing, by the computer processor configured to automate categorization, a confidence in the categorization of the current transaction as not anomalous based on the stored ranks, wherein the confidence indicates whether the current transaction for the party is a fraudulent transaction, and wherein the current transaction is an Internet log-in. - View Dependent Claims (8, 9, 10, 11, 12, 13, 14)
-
-
15. A computer-implemented method comprising:
-
for each previous transaction among one or more historical transactions for a party; determining, by a computer processor configured to automate categorization, a similarity value between a current transaction for the party and the previous transaction, wherein the similarity value is determined by computing an initial weight for each of the properties of a set of properties, computing a similarity between each of the properties of the current transaction and the properties of the previous transaction, adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity value by summing the products of the normalized adjusted weights and the computed similarities; determining, by the computer processor configured to automate categorization, that the similarity value is less than a predetermined threshold value; categorizing, by an anomaly confidence generator component of the computer processor, the current transaction as anomalous in response to determining that the similarity value is less than the predetermined threshold value; determining, by the computer processor configured to automate categorization, a factor for the previous transaction based on an age of the previous transaction; computing, by the computer processor configured to automate categorization, a rank for the previous transaction using the similarity value and the factor; and storing, by the computer processor configured to automate categorization, the computed rank in a database configured to store the computed rank; and computing, by the computer processor configured to automate categorization, a confidence in the categorization of the current transaction as anomalous based on the stored ranks, wherein the confidence indicates whether the current transaction for the party is a fraudulent transaction, and wherein the current transaction is an Internet log-in. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22)
-
-
23. A system comprising:
-
a computer processor configured to automate categorization; and a non-transitory computer readable medium coupled to the computer processor and having instructions stored thereon, which when executed by the computer processor, cause the computer processor to; for each previous transaction among one or more historical transactions for a party; determine a similarity value between a current transaction for the party and the previous transaction, wherein the similarity value is determined by computing an initial weight for each of the properties of a set of properties, computing a similarity between each of the properties of the current transaction and the properties of the previous transaction, adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity value by summing the products of the normalized adjusted weights and the computed similarities; determine that the similarity value is greater than or equal to a predetermined threshold value; categorize, via an anomaly confidence generator component of the computer processor, the current transaction as not anomalous in response to the determination that the similarity value is greater than or equal to the predetermined threshold value; determine a factor for the previous transaction based on an age of the previous transaction; compute a rank for the previous transaction using the similarity value and the factor; and store the computed rank in a database configured to store the computed rank; and compute a confidence in the categorization of the current transaction as not anomalous based on the stored ranks, wherein the confidence indicates whether the current transaction for the party is a fraudulent transaction, and wherein the current transaction is an Internet log-in. - View Dependent Claims (24, 25, 26, 27, 28)
-
Specification