Systems and methods for analyzing data
Information regarding individuals that fit a bad performance definition, such as individuals that have previously defaulted on a financial instrument or have declared bankruptcy, is used to develop a model that is usable to determine whether an individual that does not fit the bad performance definition is more likely to subsequently default on a financial instrument or to declare bankruptcy. The model may be used to generate a score for each individual, and the score may be used to segment the individual into a segment of a segmentation structure that includes individuals with related scores, where segments may include different models for generating a final risk score for the individuals assigned to the particular segments. The segment to which an individual is assigned, which may be determined based at least partly on the score assigned to the individual, may affect the final risk score that is assigned to the individual.
1. A non-transitory computer readable medium storing computer executable instructions that, when executed by one or more computer systems, configure the one or more computer systems to segment a plurality of individuals into one of a plurality of segments of a segmentation structure by performing operations comprising:
retrieving, from an electronic data store, information defining a bad performance definition, wherein the bad performance definition is defined to include individuals that have characteristics of at least one of a first failure mode of having defaulted on a financial instrument or a second failure mode of having declared bankruptcy;
obtaining first data comprising financial and demographic information regarding a plurality of individuals fitting the bad performance definition, the first data indicating characteristics of the individuals at an observation point in time;
obtaining second data comprising financial and demographic information regarding the plurality of individuals fitting the bad performance definition, the second data indicating behaviors of the individuals fitting the bad performance definition during an outcome period, the outcome period beginning after the observation point in time, the behaviors measured for the individuals fitting the bad performance definition during the outcome period include repayment performance, bankruptcy filing and response to a marketing offer;
based on both (a) the first data indicating the characteristics of the individuals at the observation point in time and (b) the second data indicating the behaviors of the individuals during the outcome period, generating a segmentation model configured to determine whether an individual not fitting the bad performance definition is more likely to enter the first failure mode or the second failure mode;
applying the segmentation model to at least a first portion of data associated with a first individual not fitting the bad performance definition, the first portion of the data comprising financial and demographic information of the first individual;
based on a first determination that an output of the segmentation model indicates that the first individual is more likely to enter the first failure mode than the second failure mode:
assigning the first individual to a first segment of a segmentation structure, the first segment comprising a first risk score model;
applying the first risk score model to the data associated with the first individual to determine a first risk score for the first individual;
allotting one or more first adverse action codes to the first individual based on relevance of the first determination on the first risk score for the first individual;
providing via a graphical user interface the first risk score to the first individual;
receiving, via the graphical user interface and responsive to the provided first risk score, a request for information that contributed to the provided first risk score; and
providing, via the graphical user interface and responsive to the request, the one or more first adverse action codes to the first individual;
applying the segmentation model to at least a second portion of data associated with a second individual not fitting the bad performance definition, the second portion of the data comprising financial and demographic information of the second individual; and
based on a second determination that the output of the segmentation model indicates that the second individual is more likely to enter the second failure mode than the first failure mode:
assigning the second individual to a second segment of the segmentation structure, the second segment comprising a second risk score model, wherein the second segment is different than the first segment;
applying the second risk score model to the data associated with the second individual to determine a second risk score for the second individual, wherein the second risk score model is different than the first risk score model; and
allotting one or more second adverse action codes to the second individual based on relevance of the second determination on the second risk score for the second individual.
2. The non-transitory computer readable medium of claim 1 , wherein the observation time comprises a plurality of months prior to generation of the model.
3. The non-transitory computer readable medium of claim 2 , wherein the outcome period is a period approximating the plurality of months but excludes the observation time.
4. The non-transitory computer readable medium of claim 1 , wherein the first segment is part of a segmentation structure comprising at least two hierarchical levels of segments.
5. The non-transitory computer readable medium of claim 4 , wherein the operations further comprise assigning the first individual to a parent segment in the segmentation structure based at least in part on output of the model.
6. The non-transitory computer readable medium of claim 1 , wherein the operations further comprise adjusting the first risk score based at least in part on the assignment of the first individual to the first segment.
7. The non-transitory computer readable medium of claim 6 , wherein the operations further comprise adjusting the second risk score based at least in part on the assignment of the second individual to the second segment.
8. A computer-implemented method to segment a plurality of individuals into one of a plurality of segments of a segmentation structure, the method comprising:
retrieving, from an electronic data store, information defining a bad performance definition, wherein the bad performance definition is defined to include individuals that have characteristics of at least one of a first failure mode of having defaulted on a financial instrument or a second failure mode of having declared bankruptcy;
obtaining first data comprising financial and demographic information regarding a plurality of individuals fitting the bad performance definition, the first data indicating characteristics of the individuals at an observation point in time;
obtaining second data comprising financial and demographic information regarding the plurality of individuals fitting the bad performance definition, the second data indicating behaviors of the individuals fitting the bad performance definition during an outcome period, the outcome period beginning after the observation point in time, the behaviors measured for the individuals fitting the bad performance definition during the outcome period include repayment performance, bankruptcy filing and response to a marketing offer;
based on both (a) the first data indicating characteristics of the individuals at the point in observation time and (b) the second data indicating the behaviors of the individuals during the outcome period, generating a segmentation model configured to determine whether an individual not fitting the bad performance definition is more likely to enter the first failure mode or the second failure mode;
applying the segmentation model to at least a first portion of data associated with a first individual not fitting the bad performance definition, the first portion of the data comprising financial and demographic information of the first individual;
based on a first determination that an output of the segmentation model indicates that the first individual is more likely to enter the first failure mode than the second failure mode:
assigning the first individual to a first segment of a segmentation structure, the first segment comprising a first risk score model;
applying the first risk score model to the data associated with the first individual to determine a first risk score for the first individual;
allotting one or more first adverse action codes to the first individual based on relevance of the first determination on the first risk score for the first individual;
providing via a graphical user interface the first risk score to the first individual;
receiving, via the graphical user interface and responsive to the provided first risk score, a request for information that contributed to the provided first risk score; and
providing, via the graphical user interface and responsive to the request, the one or more first adverse action codes to the first individual;
applying the segmentation model to at least a second portion of data associated with a second individual not fitting the bad performance definition, the second portion of the data comprising financial and demographic information of the second individual; and
based on a second determination that the output of the segmentation model indicates that the second individual is more likely to enter the second failure mode than the first failure mode:
assigning the second individual to a second segment of the segmentation structure, the second segment comprising a second risk score model, wherein the second segment is different than the first segment;
applying the second risk score model to the data associated with the second individual to determine a second risk score for the second individual, wherein the second risk score model is different than the first risk score model; and
allotting one or more second adverse action codes to the second individual based on relevance of the second determination on the second risk score for the second individual.
9. The computer-implemented method of claim 8 , wherein the first segment is part of a segmentation structure comprising at least two hierarchical levels of segments.
10. The computer-implemented method of claim 9 , further comprising assigning the first individual to a parent segment in the segmentation structure based at least in part on output of the model.
11. The computer-implemented method of claim 8 , further comprising adjusting the first risk score based at least in part on the assignment of the first individual to the first segment.
12. The computer-implemented method of claim 8 , wherein the first failure mode comprises filing for bankruptcy and the second failure mode comprises defaulting on a financial instrument.
13. The computer-implemented method of claim 8 , wherein the first failure mode comprises defaulting on an installment loan and the second failure mode comprises defaulting on a revolving loan.
14. The computer-implemented method of claim 8 , wherein the first failure mode comprises defaulting on a bank loan and the second failure mode comprises defaulting on an automobile loan.
15. A computer system to segment a plurality of individuals into one of a plurality of segments of a segmentation structure, the computer system comprising:
memory; and
a hardware processor in communication with the memory and configured with processor-executable instructions to perform operations comprising:
obtaining a bad performance definition, wherein the bad performance definition is defined to include individuals that have characteristics of at least one of a first failure mode of having defaulted on a financial instrument or a second failure mode of having declared bankruptcy;
obtaining first data comprising financial and demographic information regarding a plurality of individuals fitting the bad performance definition, the first data indicating characteristics of the individuals at an observation point in time;
obtaining second data comprising financial and demographic information regarding the plurality of individuals fitting the bad performance definition, the second data indicating behaviors of the individuals fitting the bad performance definition during an outcome period, the outcome period beginning after the observation point in time, the behaviors measured for the individuals fitting the bad performance definition during the outcome period include repayment performance, bankruptcy filing and response to a marketing offer;
based on both (a) the first data indicating the characteristics of the individuals at the observation point in time and (b) the second data indicating the behaviors of the individuals during the outcome period, generating a segmentation model configured to determine whether an individual not fitting the bad performance definition is more likely to enter the first failure mode or the second failure mode;
applying the segmentation model to at least a first portion of data associated with a first individual not fitting the bad performance definition, the first portion of the data comprising financial and demographic information of the first individual;
based on a first determination that an output of the segmentation model indicates that the first individual is more likely to enter the first failure mode than the second failure mode:
assigning the first individual to a first segment of a segmentation structure, the first segment comprising a first risk score model;
applying the first risk score model to the data associated with the first individual to determine a first risk score for the first individual;
allotting one or more first adverse action codes to the first individual based on relevance of the first determination on the first risk score for the first individual;
providing via a graphical user interface the first risk score to the first individual;
receiving, via the graphical user interface and responsive to the provided first risk score, a request for information that contributed to the provided first risk score; and
providing, via the graphical user interface and responsive to the request, the one or more first adverse action codes to the first individual;
applying the model to at least a second portion of data associated with a second individual not fitting the bad performance definition, the second portion of the data comprising financial and demographic information of the second individual; and
based on a second determination that the output of the segmentation model indicates that the second individual is more likely to enter the second failure mode than the first failure mode:
assigning the second individual to a second segment of the segmentation structure, the second segment comprising a second risk score model, wherein the second segment is different than the first segment;
applying the second risk score model to the data associated with the second individual to determine a second risk score for the second individual, wherein the second risk score model is different than the first risk score model; and
allotting one or more second adverse action codes to the second individual based on relevance of the second determination on the second risk score for the second individual.
16. The computer system of claim 15 , wherein the first data comprises demographic data and financial data regarding the plurality of individuals.
17. The computer system of claim 15 , wherein the operations further comprise adjusting the first risk score based at least in part on the assignment of the first individual to the first segment.
18. The computer system of claim 15 , wherein the operations further comprise assigning the first individual to a parent segment of the first segment in a segmentation structure based at least in part on output of the model.