DATA ANALYTICS MODELS FOR LOAN TREATMENT

• US 20110270779A1
• Filed: 04/30/2010
• Published: 11/03/2011
• Est. Priority Date: 04/30/2010
• Status: Active Grant

• Current Assignee: Corelogic Solutions LLC (CoreLogic Incorporated)
• Original Assignee: Corelogic Solutions LLC (CoreLogic Incorporated)
• Inventors: Showalter, Thomas
• Granted Patent: US 8,458,074 B2
• Time in Patent Office: Days
• US Class Current: 705/36.R
• CPC Class Codes: G06Q 40/02 Banking, e.g. interest calculation or account maintenance credit or loans G06Q40/03
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof
• CPC Class Codes: G06Q 40/06 Asset management; Financial planning or analysis

Abstract

Data analytics are provided in loan treatment. Various sources of data may be used to optimize or predict value for a loan. Using machine-learning and/or statistical analysis, loans or treatment best suited for a particular borrower may be determined. Due to the large amounts of data available, borrower behavior may be learned from previous behavior of others and mapped to a predictive model. Machine-learning indicates the most relevant factors in loan treatment, providing a matrix for predicting loan value or treatment success. A given borrower may be classified into one of many classes of borrower based on credit information, property information, desired loan information, real estate market information, and/or other data. Tens, hundreds, or even thousands of variables may be used to predict the optimum treatment.

Claims

1. A system for data analytics in loan treatment, the system comprising:

an input configured to receive credit report information for a person associated with a loan, property information for a specific property associated with the loan, loan information for the loan, and real estate market information for a region including the property; and

a processor configured to apply;

a cluster model comprising an unsupervised machine-learned classifier configured to classifying a borrower of the loan and the specific property into one of a plurality of borrower-property clusters, each of the borrower-property clusters being a function of both the credit report information and the property information;

a treatment model comprising a supervised machine-learned classifier configured to output a loan treatment recommendation for the borrower of the loan, the processor configured to apply the treatment model as a function of the one of the borrower-property clusters, the property information, the loan information, and the real estate market information, the loan treatment recommendation selected from a plurality of possible treatments as one of the possible treatments with a greater value over a period of time, the possible treatments including loan modification and short sale.

2. The system of claim 1 further comprising a display configured to output the loan treatment recommendation.

3. A system for data analytics in loan treatment, the system comprising:

an input configured to receive credit report information for a person associated with a loan, property information for a specific property associated with the loan, loan information for the loan, and real estate market information for a region including the property, the real estate market information including a time on market for a region associated with the specific property, the property information including a property value, equity, and open lean amount;

a processor configured to extract metrics including a willingness, a capacity and a stress from the credit information of a borrower of the loan, and configured to apply a model, the model comprising a machine-learned classifier configured to output a loan treatment recommendation for the borrower of the loan, the processor configured to apply the model as a function of the willingness, the capacity, the stress, the time on market, the property value, the equity, the open lean amount, a combined loan-to-value ratio, the loan information, and the real estate market information; and

a display configured to output the loan treatment recommendation.

4. The system of claim 3 wherein the input comprises a network interface configured to receive one or more files including the credit report information, the processor extracting the willingness, the capacity, and the stress from the one or more files.

5. A system for data analytics in loan treatment, the system comprising:

an input configured to receive credit report information for a person associated with a loan, property information for a specific property associated with the loan, loan information for the loan, and real estate market information for a region including the property;

a processor configured to apply a model, the model comprising a machine-learned classifier configured to output a loan treatment recommendation for a borrower of the loan, the processor configured to apply the model as a function of the credit report information, the property information, the loan information, and the real estate market information; and

a display configured to output the loan treatment recommendation.

6. The system of claim 5 wherein options for the loan treatment recommendation comprise one of modification of the loan, foreclosure of the loan, short sale, and forbearance, the model configured to output one of the options as a function of the credit report information, the property information, the loan information, and the real estate market information.

7. The system of claim 5 wherein the credit report information, the property information, the loan information, and the real estate market information comprise metrics, at least 90% of the metrics used by the model being from the credit report information.

8. The system of claim 5 wherein the credit report information comprises metrics, the processor configured to determine distress, capacity to pay, and willingness to pay factors from the metrics, the model using the distress, capacity to pay, and willingness to pay factors as input feature vectors.

9. The system of claim 5 wherein the model comprises a cluster classifier configured to cluster classify the borrower as belonging to one of a plurality of clusters based on the credit report information and the property information.

10. The system of claim 9 wherein the plurality of clusters comprise at least ten clusters, wherein the model comprises at least two separate loan treatment classifiers for each of the clusters, the at least two separate loan treatment classifiers outputting a net present value of the property as a function of the loan information and the real estate information, the loan treatment recommendation output by the model comprising the loan treatment recommendation associated with the highest net present value of the property output by the separate loan treatment classifiers for the cluster to which the borrower belongs.

11. The system of claim 5 wherein the input is configured to receive a value for a loan retention variable, the processor configured to apply the model as a function of the value of the loan retention variable, the loan treatment recommendation being a function of the value.

12. The system of claim 5 wherein the processor is configured to calculate a plurality of credit factors for each of a plurality of types of debt from the credit report information, the model applying the credit factors as input vectors.

13. The system of claim 5 further comprising a user selector, the user selector configured to indicate selection of one or more values of variables for the loan information, the loan treatment recommendation being a function of the one or more values such that a user interacts with the model to receive different outputs.

14. The system of claim 5 wherein the input comprises a network interface configured to receive the credit report information;

further comprising a database memory having stored therein the real estate market information;

wherein the processor is configured to extract values for at least one thousand metrics from the credit report information, the property information, the loan information, and the real estate information, a majority of the metrics being from the credit report information, the processor configured to calculate credit factors from the credit report information, calculate product factors from the loan information, and calculate property factors from the property information, the credit factors and product factors being input feature vectors for the model, the model determining a behavior cluster for the borrower from the input feature vectors.

15. In a computer readable storage medium having stored therein data representing instructions executable by a programmed processor for data analytics in loan processing, the instructions comprising:

modeling borrower loan behavior from credit values for a borrower;

modeling property behavior from property characteristics for a property associated with the borrower;

classifying the borrower into one of a plurality of clusters based on the modeled borrower loan behavior and the modeled property behavior; and

outputting information based on the one cluster to which the borrower is classified.

16. The computer readable storage medium of claim 15 wherein modeling borrower loan behavior comprises calculating a distress factor, a capacity factor, and a willingness to pay factor from the credit values, the distress, capacity, and willingness factors each comprising one or more metrics, the metrics input as a feature vector for the classifying.

17. The computer readable storage medium of claim 16 wherein the distress factor comprises an indication of debt servicing distress of the borrower and is calculated as a function of trade lines for revolving credit, auto credit, finance company credit, or combinations thereof.

18. The computer readable storage medium of claim 15 wherein modeling the property behavior comprises calculating a property value factor, a combined loan to Value factor, and an appreciation factor from the property characteristics and real estate market information, the property value, combined loan to value, and appreciation factors each comprising one or more metrics, the metrics input as a feature vector for the classifying.

19. The computer readable storage medium of claim 15 further comprising:

determining a percent chance of default as a function of the one cluster.

20. The computer readable storage medium of claim 15 further comprising:

outputting a loan treatment recommendation as a function of the classification of the borrower into the one cluster, the loan treatment recommendation comprising one of loan modification, short sale, and foreclosure, the loan modification output where the borrower loan behavior is modeled as able and willing to execute, the short sale output where the property behavior is modeled as able to support short sale, and otherwise outputting foreclosure.

21. A method for data analytics in loan processing, the method comprising:

generating, with a computer, a user interface;

receiving, with a user input of the computer, an indication of a property value, a loan term, a loan modification, or combinations thereof, the indication received as part of the user interface;

calculating, with the computer, a net present value as a function of a machine-trained model, the received indication being input as a feature of the machine-trained model;

repeating the receiving and calculating for different indications for a same loan of a same property;

displaying on a display the net present value for each of the different indications.

22. The method of claim 21 wherein calculating comprises determining a borrower feature vector of metrics from credit data, determining a property feature vector of metrics from data specific to the property and from real estate data for a region associated with the property, determining a loan feature vector for loan terms of the loan, and inputting the borrower, property, and loan feature vectors into the machine-trained model, the net present value determined as a function of an output of the machine-trained model.

23. The method of claim 22 wherein the borrower feature vector includes at least one hundred metrics extracted from a credit report.

24. The method of claim 21 wherein receiving comprises receiving an indication of a level of retention for the loan.

25. In a computer readable storage medium having stored therein data representing instructions executable by a programmed processor for data analytics in loan processing, the instructions comprising:

applying a machine trained model to borrower and property data for a distressed loan;

calculating net present value as a function of an output of the application of the machine trained model, the net present value calculated for each of a plurality of loan treatments; and

outputting the net present value of at least one of the loan treatments.

26. The computer readable storage medium of claim 25 further comprising:

repeating the calculating for different terms for the loan treatments.

27. The computer readable storage medium of claim 25 wherein the calculating comprises using a Bayesian classifier

28. The computer readable storage medium of claim 25 wherein outputting comprises outputting a first of the loan treatments associated with a highest net present value.

29. The computer readable storage medium of claim 28 wherein the highest net present value is skewed as a function of a level of loan retention policy.

30. A method for data analytics in loan processing, the method comprising:

receiving, with a computer, a portfolio of loans;

extracting, with the computer, borrower and property information for the loans of the portfolio;

applying, with the computer, a machined-trained model estimating survivability, for each of the loans of the portfolio, of a plurality of loan treatments;

selecting, with the computer, loan treatments for the loans as a function of the survivability;

calculating, with the computer, a cost for the portfolio as a function of the selected loan treatments; and

outputting the cost and the selected loan treatments.

31. The method of claim 30 wherein calculating the cost comprises calculating a net present value for the selected loan treatments.

32. The method of claim 30 wherein selecting comprises selecting as a function of a number or percentage of one of the loan treatments available for the portfolio.

33. In a computer readable storage medium having stored therein data representing instructions executable by a programmed processor for data analytics in loan processing, the instructions comprising:

determining at least twenty borrower metrics from credit report information for first and second borrowers, the borrower metrics including distress, capacity, and willingness metrics;

determining at least eight property metrics about a property associated with first and second loans for the first and second borrowers, respectively, the property metrics including property value, equity, and combined loan-to-value ratio metrics;

classifying the first borrower into a first cluster of a plurality of clusters based on the at least twenty borrower metrics and the at least eight property metrics for the first borrower, the metrics for the classifying including the distress, capacity, willingness, property value, equity, and combined loan-to-value ratio metrics for the first borrower;

classifying the second borrower into a second cluster of a plurality of clusters based on the at least twenty borrower metrics and the at least eight property metrics for the second borrower, the metrics for the classifying including the distress, capacity, willingness, property value, equity, and combined loan-to-value ratio metrics for the second borrower;

wherein the first cluster corresponds to a greater default rate than the second cluster with the equity of the first borrower being higher than the equity of the second borrower; and

recommending foreclosure for the first borrower and a different loan treatment for the second borrower based on the classification into the first and second clusters, respectively.