Risk determination and management using predictive modeling and transaction profiles for individual transacting entities

A computer implemented method of determining a level of risk for a transaction in an account of a transacting entity, the method comprising:

• for each of a first plurality of transacting entities, generating a profile of transaction patterns of the transacting entity using historical transactions of the transacting entity;
• generating and storing a predictive model of fraudulent transactions from a plurality of fraudulent transactions and the profiles of the transacting entities of the fraudulent transactions, and a plurality of non-fraudulent transactions and the profiles of the transacting entities of the non-fraudulent transactions;
• receiving in real time a current transaction of a transacting entity for a type of account, the current transaction received prior to completion of the transaction by the transacting entity;
• generating in real time a signal indicative of the level of fraud associated with the current transaction by applying the current transaction and a profile summarizing a pattern of historical transactions of the transacting entity to the predictive model;
• and
• transmitting in real time the signal indicative of the level of fraud to at least one of the transacting entity or a second entity to allow for either completion or termination of the transaction.

A computer implemented method of determining a level of risk for a transaction in an account of a transacting entity, the method comprising:

• receiving in real time, data pertaining to a pending first transaction of a transacting entity for a type of account, the data received prior to completion of the first transaction by the transacting entity;
• generating in real time a signal indicative of the level of risk associated with the first transaction by comparing the data pertaining to the first transaction of the transacting entity with a profile summarizing a pattern of historical transactions of the transacting entity;
• determining in real time an authorization response for the first transaction as a function of the signal indicative of risk associated with the transaction;
• transmitting in real time the authorization response to either the transacting entity or a second entity to allow either the transacting entity or the second entity to terminate or complete the first transaction;
• and
• determining whether to approve or decline a second transaction as a function of the signal indicative of the level of risk associated with the first transaction.

A computer implemented method of determining a fraud score for a transaction in an account of a transacting entity, the method comprising:

• generating a predictive model of risk-associated transactions from high risk and low risk historical transactions of transacting entities and profiles of the transacting entity, for substantially the same type of account as the transaction by:
• selecting high risk and low risk transactions and transacting entity data associated with the selected transactions;
• segregating the selected transactions into account-days, with each account-day representing at least one transaction of the accound during a day;
• randomly selecting low risk account-days, and for any sequence of high risk account-days of an individual account, selecting only initial high risk account-day;
• and
• generating risk related variables from the randomly selected low risk account days and the selected initial high risk account days and the profiles associated with the selected accounts;
• for each of a plurality of transactions in a batch of transactions, generating a fraud score for the transaction by applying the transaction and a profile of the transacting entity of the transaction to the predictive model;
• authorizing each of the transactions in the batch regardless of the fraud score associated with the transaction;
• and
• for each transaction in the batch, responsive to the fraud score of the transaction exceeding a threshold amount, designating the account associated with the transacting entity of the transaction as a high risk account.

A computer implemented method of determining a fraud score for a transaction in an account of a transacting entity, the method comprising:

• storing a predictive model of risk-associated transactions generated from high risk and low risk historical transactions that are for substantially the same type of account as current transactions of transacting entities and profiles of the transacting entity, each profile summarizing a pattern of historical transactions of a transacting entity, for substantially the same type of account as the transaction;
• receiving in real time a current transaction of a transacting entity, the current transaction received prior to completion of the transaction by the transacting entity;
• authorizing the current transaction prior to generating a fraud score associated with the current transaction;
• generating a continuous fraud score indicative of a probability that the current transaction is fraudulent by applying the current transaction of the transacting entity and the profile of the transacting entity to the predictive model;
• and
• responsive to the fraud score of the current transaction exceeding a threshold amount, designating the account associated with the transacting entity of the current transaction as a high risk account so as to prevent subsequent transactions of the transacting entity from being authorized.

A computer implemented method for developing a predictive model for determining a risk score indicative of a level of risk in a transaction associated with a transacting entity, comprising the operations of:

• receiving for a plurality of transacting entities, historical transaction data for transactions occurring over a period of time;
• for each of the transacting entities, creating a profile summarizing patterns of the transactions of the transacting entity;
• and
• creating the predictive model using the transaction data of each transaction for a type of account, the profile of the transacting entity making each transaction wherein the profile is for substantially the same type of account as each transaction, and data categorizing each transaction with respect to a level of risk in the transaction, wherein creating the predictive model includes;
• selecting high risk transactions and accounts associated therewith, and a random sample of low risk transactions and accounts associated therewith;
• segregating the selected transactions into account-days, each account-daylassociated with one of the selected accounts and at least one transaction of the account occurring during a given day;
• selecting account-days by randomly selecting account-days that do not include a high risk transaction, and for any sequence of account-days in an account each of which include at least one high risk transaction, selecting only an earliest account-day;
• and
• generating risk-related variables from the selected account-days and the profiles associated with the selected accounts.

A computer program product for controlling a computer system to process transactions in accounts of account holders, comprising:

• program code communicatively coupled to a database of account information for account holders and account transactions for the account holders that generates a profile of an account holder by summarizing patterns of historical transactions of the account holder for a type of account;
• program code that communicatively couples with a communications network and receives in real time a current transaction of an account holder from a point of sale device coupled to the communications network and performing the current transaction, wherein the current transaction is for substantially the same type of account as the historical transactions;
• program code that receives in real time data from the current transaction and generates a signal indicative of a level of risk associated with the current transaction by comparing the current transaction of the account holder with the profile of the account holder;
• program code that couples with the generating program code to receive the signal indicative of the level of risk and that generates in real time an authorization response signal for the current transaction as a function of the level of risk;
• program code that transmits in real time the authorization signal to the point of sale device;
• program code that updates the profile of the account holder using the current transaction;
• and
• a computer readable medium that stores the program codes.

A computer program product for controlling a computer system to process transactions in accounts of account holders, comprising:

• program code communicatively coupled to a database of account information for account holders and account transactions for the account holders that generates a profile of an account holder by summarizing patterns of historical transactions of the account holder for a type of account;
• program code that communicatively couples with a communications network and receives in real time a current transaction of an account holder from a point of sale device coupled to the communications network and performing the current transaction, wherein the current transaction is for substantially the same type of account as the historical transactions;
• program code that receives in real time data from the current transaction and generates a signal indicative of a level of risk associated with the current transaction by comparing the current transaction of the account holder with the profile of the account holder;
• program code that couples with the generating program code to receive the signal indicative of the level of risk and that generates in real time an authorization response signal for the current transaction as a function of the level of risk;
• program code that transmits in real time the authorization signal to the point of sale device;
• program code that stores a predictive model of risk associated transactions developed from the historical transactions of the account holders and the profiles of the account holders;
• program code that compares the current transaction of the account holder to the profile of the account holder by applying the current transaction and the profile of the account holder to the predictive model;
• and
• a computer readable medium that stores the projram codes.

A computer program product for controlling a computer system to determine a level of risk in an account of a transacting entity, the program product comprising:

• a computer readable medium that stores program code;
• a database including, for each of a plurality of transacting entities, a profile of historical transaction patterns of the transacting entity for a type of account;
• program code for executing a predictive model of risk associated transactions, the predictive model generated from high risk and low risk transactions and profiles of transacting entities;
• program code for receiving in real time a current transaction of a transacting entity, wherein the current transaction is for substantially the same type of account as the historical transactions;
• program code for generating in real time a signal indicative of the level of risk associated with the current transaction by applying the current transaction and the profile of the transacting entity to the predictive model;
• program code for determining in real time whether to approve or decline the current transaction according to the signal indicative of risk associated with the current transaction;
• and
• program code, responsive to the level of risk of the current transaction exceeding a threshold amount, for not authorizing the current transaction;
• program code, responsive to the level of risk of the current transaction exceeding the threshold amount, for designating the account associated with the transacting entity of the current transaction as a high risk account so as to prevent subsequent transactions of the transacting entity from being authorized.

A computer program product for controlling a computer system to process accounts associated with transactions, the program product comprising:

• a computer readable medium that stores program code;
• program code that receives a plurality of current transactions for a plurality of transacting entities, each current transaction associated with an account of a transacting entity;
• program code that generates a continuous fraud score for each current transaction indicating a likelihood that the current transaction is fraudulent;
• program code that ranks the accounts by the fraud scores of their respective current transactions, from a most significant fraud score to a least significant fraud score;
• and
• program code that provides user selectable fraud control actions for application to selected ones of the accounts.

A computer program product for controlling a computer system to develop a predictive model for determining a risk score indicative of a level of risk in a transaction associated with a transacting entity, comprising:

• a computer readable medium that stores program code;
• program code that receives for each of a plurality of transacting entities, historical transaction data for transactions of the transacting entity occurring over a period of time;
• program code that creates, for each of the transacting entities, a profile summarizing patterns of the transactions of the transacting entity;
• and
• program code that creates the predictive model using the transaction data of each transaction for a type of account, the profile of the transacting entity making each transaction wherein the profile is for substantially the same type of account as each transaction, and data categorizing each transaction with respect to a level of risk in the transaction, wherein program code that creates the predictive model further comprises program code that:
• selects high risk transactions and accounts associated therewith, and a random sample of low risk transactions and accounts associated therewith;
• segregates the selected transactions into time intervals, each time interval associated with one of the selected accounts and at least one transaction of the account during the time interval;
• randomly selects time intervals that do not include a high risk transaction, and for any sequence of time intervals each including at least one high risk transaction, selects only and at least one of an earliest time interval;
• and
• generates risk-related variables from the selected time intervals and the profiles associated with the selected accounts.

A user interface of a computer program product that executes on a computer system for detecting fraudulent transactions, the user interface comprising:

• an account number display field for displaying an account number of an account holder;
• an account holder name display field for displaying the name of the account holder;
• a fraud score display field for displaying a fraud score indicating the likelihood that a current transaction for the account holder is fraudulent;
• a display field for displaying at least one transaction of the account holder having a current transaction with a fraud score indicating that it is likely that the current transaction is fraudulent;
• and
• at least one reason display field for displaying a reason the current transaction is likely to be fraudulent.

A computer assisted method of processing a credit card transaction, the method comprising:

• receiving in real time a current credit card transaction of a credit card account holder;
• deriving variables related to the current credit card transaction of the credit card account holder and past credit card transactions of the credit card account holder by:
• accessing a profile of the credit card account holder, the profile summarizing past credit card transactions of the credit card account holder;
• and
• deriving some of the variables from the profile;
• and
• generating in real time from the derived variables a continuous fraud score indicating a likelihood that the current credit card transaction is fraudulent.