Predictive modeling of consumer financial behavior using supervised segmentation and nearest-neighbor matching

• US 7,165,037 B2
• Filed: 12/14/2004
• Issued: 01/16/2007
• Est. Priority Date: 05/06/1999
• Status: Expired

• Current Assignee: Callahan Cellular LLC (Intellectual Ventures LLC)
• Original Assignee: Fair Isaac Corporation
• Inventors: Dunning, Ted E., Russell, Gerald R., Sitze, Kevin L., Lazarus, Michael A., Peranich, Larry S., Vernhes, Frederique, Blume, A. U. Mattias, Brown, Kenneth B., Caid, William R.
• Primary Examiner(s): Jeanty, Romain
• Assistant Examiner(s): Sterrett, Jonathan G.
• Application Number:
  US11/012,812
• Publication Number:
  US 20050159996A1
• Time in Patent Office: 763 Days
• Field of Search: 705/10
• US Class Current: 705/7.31
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0207 Discounts or incentives, e.g. coupons or rebates
• CPC Class Codes: G06Q 30/0241 Advertisements
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute
• CPC Class Codes: G06Q 30/0601 Electronic shopping [e-shopping]

Abstract

Predictive modeling of consumer financial behavior, including determination of likely responses to particular marketing efforts, is provided by application of consumer transaction data to predictive models associated with merchant segments, which are derived from the consumer transaction data based on co-occurrences of merchants in sequences of transactions. Merchant vectors represent specific merchants, and are aligned in a vector space as a function of the degree to which the merchants co-occur. Supervised segmentation is applied to merchant vectors to form merchant segments. Merchant segment predictive models provide predictions of spending in each merchant segment for any particular consumer, based on previous spending by the consumer. Consumer profiles describe summary statistics of each consumer'"'"'s spending in the merchant segments, and across merchant segments. Consumer profiles include consumer vectors derived as summary vectors of selected merchants patronized by the consumer. Predictions of consumer behavior are made by applying nearest-neighbor analysis to consumer vectors.

Claims

1. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein identifying at least one nearest neighbor comprises identifying consumer vectors having a dot product between the consumer vector and the target consumer vector that exceeds a predetermined threshold.

2. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein identifying at least one nearest neighbor comprises identifying a predetermined number of consumer vectors having the highest dot products between the consumer vector and the target consumer vector.

3. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

further comprising fusing the generated behavior prediction with additional data to generate a second-level behavior prediction.

4. The method of claim 3, wherein fusing comprises:

training a second-level predictive model with generated behavior predictions and additional data; and

applying the generated behavior prediction and additional data to the trained second-level predictive model to obtain a second-level behavior prediction.

5. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the consumer vectors for the reference set exclude target product purchases.

6. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the reference set of consumers is selected randomly.

7. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

for a reference set of consumers, obtaining consumer vectors and data describing financial behavior;

obtaining a consumer vector for the target consumer;

identifying at least one nearest neighbor to the target consumer vector among the reference set of consumers; and

generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the reference set of consumers is selected non-randomly, and further comprising adjusting the generated behavior prediction to compensate for the non-randomness of the reference set selection.

8. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

generating consumer vectors for a plurality of consumers;

defining at least one consumer segment having predicted financial behavior data;

determining a consumer segment for the target consumer; and

based on the determined consumer segment, generating predicted financial behavior for the target consumer;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein defining at least one consumer segment comprises;

initializing a set of consumer segment vectors;

accepting at least one segment label for at least one of the consumers;

for each of at least a subset of the labeled consumers;

selecting at least one consumer segment vector for a consumer;

determining whether the selected consumer segment vector matches the segment label for the consumer; and

responsive to the determination, adjusting zero or more of the consumer segment vectors.

9. A computer implemented method of predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

generating consumer vectors for a plurality of consumers;

defining at least one consumer segment having predicted financial behavior data;

determining a consumer segment for the target consumer; and

based on the determined consumer segment, generating predicted financial behavior for the target consumer;

wherein generating a behavior prediction comprises;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the target consumer is associated with a target consumer vector, and wherein determining a consumer segment for the target consumer comprises selecting a consumer segment corresponding to a consumer segment vector having the highest dot product between the consumer segment vector and the target consumer vector.

10. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the merchant segment predictive model identifies at least one nearest neighbor by identifying consumer vectors having a dot product between the consumer vector and the target consumer vector that exceeds a predetermined threshold.

11. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the merchant segment predictive model identifies at least one nearest neighbor by identifying a predetermined number of consumer vectors having the highest dot products between the consumer vector and the target consumer vector.

12. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the merchant segment predictive model fuses the generated behavior prediction with additional data to generate a second-level behavior prediction.

13. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the consumer vectors for the reference set exclude target product purchases.

14. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the reference set of consumers is selected randomly.

15. A system for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

a merchant vector build module, for generating one or more merchant vectors for at least a subset of merchants;

a segmentation module for applying segmentation to said one or more merchant vectors to provide at least one merchant segment;

an input device for obtaining, for a reference set of consumers, consumer vectors and data describing financial behavior; and

at least one merchant segment predictive model, corresponding to said at least one merchant segment, said model coupled to the build module and the input device, for identifying at least one nearest neighbor consumer vector to a target consumer vector among the reference set consumer vectors, and generating a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

said system configured for;

training a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

using an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

applying the predictive model to the consumer vector of the target consumer to output for said target consumer a predicted spending amount; and

wherein the reference set of consumers is selected non-randomly, and wherein the predictive model adjusts the generated behavior prediction to compensate for the non-randomness of the reference set selection.

16. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the computer-readable code adapted to identify at least one nearest neighbor comprises computer-readable code adapted to identify consumer vectors having a dot product between the consumer vector and the target consumer vector that exceeds a predetermined threshold.

17. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the computer-readable code adapted to identify at least one nearest neighbor comprises computer-readable code adapted to identify a predetermined number of consumer vectors having the highest dot products between the consumer vector and the target consumer vector.

18. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

further comprising computer-readable code adapted to fuse the generated behavior prediction with additional data to generate a second-level behavior prediction.

19. The computer-readable medium of claim 18, wherein the computer-readable code adapted to fuse comprises:

computer-readable code adapted to train a second-level predictive model with generated behavior predictions and additional data; and

computer-readable code adapted to apply the generated behavior prediction and additional data to the trained second-level predictive model to obtain a second-level behavior prediction.

20. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output and predicted spending amount for said target consumer; and

wherein the consumer vectors for the reference set exclude target product purchases.

21. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the reference set of consumers is selected randomly.

22. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to, for a reference set of consumers, obtain consumer vectors and data describing financial behavior;

computer-readable code adapted to obtain a consumer vector for the target consumer;

computer-readable code adapted to identify at least one nearest neighbor to the target consumer vector among the reference set consumer vectors; and

computer-readable code adapted to generate a financial behavior prediction for the target consumer by aggregating the financial behavior data of the consumers corresponding to the identified consumer vectors;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the reference set of consumers is selected non-randomly, and further comprising computer-readable code adapted to adjust the generated behavior prediction to compensate for the non-randomness of the reference set selection.

23. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to generate consumer vectors for a plurality of consumers;

computer-readable code adapted to define at least one consumer segment having predicted financial behavior data;

computer-readable code adapted to determine a consumer segment for the target consumer; and

computer-readable code adapted to, based on the determined consumer segment, generate predicted financial behavior for the target consumer;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the computer-readable code adapted to define at least one consumer segment comprises;

computer-readable code adapted to initialize a set of consumer segment vectors;

computer-readable code adapted to accept at least one segment label for at least one of the consumers;

computer-readable code adapted to, for each of at least a subset of the labeled consumers;

select at least one consumer segment vector for a consumer;

determine whether the selected consumer segment vector matches the segment label for the consumer; and

responsive to the determination, adjust zero or more of the consumer segment vectors.

24. A computer-readable medium comprising computer-readable code for predicting financial behavior of a target consumer with respect to an offer or merchant, comprising:

computer-readable code adapted to generate consumer vectors for a plurality of consumers;

computer-readable code adapted to define at least one consumer segment having predicted financial behavior data;

computer-readable code adapted to determine a consumer segment for the target consumer; and

computer-readable code adapted to, based on the determined consumer segment, generate predicted financial behavior for the target consumer;

wherein the computer-readable code adapted to generate a behavior prediction comprises;

computer-readable code to train a predictive model using a plurality of consumer vectors, corresponding financial behavior data, and merchant vectors;

computer-readable code to use an unexpected deviation learning approach to determine values of the merchant vectors;

wherein said unexpected deviation learning approach comprises comparing co-occurences of merchant descriptions in said financial behavior data to determine if a pair of merchants are either positively or negatively concurrent wherein either the positive or negative concurrency is used to determine values for the merchant vectors; and

computer-readable code to apply the predictive model to the consumer vector of the target consumer and output a predicted spending amount for said target consumer; and

wherein the target consumer is associated with a target consumer vector, and wherein the computer-readable code adapted to determine a consumer segment for the target consumer comprises computer-readable code adapted to select a consumer segment corresponding to a consumer segment vector having the highest dot product between the consumer segment vector and the target consumer vector.