Intelligent collections models

Abstract

Apparatuses, computer media, and methods for analyzing credit and tax form data and determining a collection treatment type to collect revenue. A collections model is constructed to determine a collections score that is based on raw credit data and tax form data and is indicative of a debtor's propensity to pay an owed amount. The collections model includes score bands, each score band being associated with a range of credit scores. A collections score is determined from a scoring expression that is associated with a score band and that typically includes a subset of available raw credit data and tax form data. A collections treatment type is determined from a collections score. Each treatment type corresponds to a treatment action that is directed to the debtor. A collections model is constructed from historical tax data, in which score bands and scoring expressions are constructed for the collections model.

1 . (canceled) 
     
     
         2 . A computer-implemented method comprising:
 generating training data for training a predictive model to estimate a likelihood of a debtor to pay a debt, the training data including, for each of a plurality of debtors, (i) historic tax return data for the debtor, (ii) commercial financial or credit data that is usable to calculate a credit score for the debtor, and (iii) a label indicating a likelihood of the debtor to pay a debt;   training the predictive model using the historic tax return data, the commercial financial or credit data, and the labels included in the training data;   after training the predictive model, identifying a particular debtor;   obtaining (i) historic tax return data for the particular debtor, and (ii) commercial financial or credit data that is usable to calculate a credit score for the particular debtor;   providing, to the predictive model, (i) the historic tax return data for the particular debtor, and (ii) the commercial financial or credit data that is usable to calculate a credit score for the particular debtor; and   obtaining, from the predictive model, an indication of a likelihood of the particular debtor to pay a debt.   
     
     
         3 . The method of  claim 2 , wherein the historic tax return data comprises one or more values were entered on a personal income tax form by an associated debtor. 
     
     
         4 . The method of  claim 2 , wherein the debt comprises a debt owed to a government revenue agency. 
     
     
         5 . The method of  claim 2 , wherein the historic tax return data comprises data that is available only to a government revenue agency or to designated agents of the government revenue agency. 
     
     
         6 . The method of  claim 2 , wherein the historic tax return data comprises data that is not reflected on a credit report of an associated debtor. 
     
     
         7 . The method of  claim 2 , wherein the historic tax return data comprises one or more values that reflect an amount of tax listed as due on a tax return in relation to an amount of income listed on the tax return. 
     
     
         8 . The method of  claim 2 , wherein the historic tax return data comprises one or more values that characterize a status of a previous year's tax return. 
     
     
         9 . The method of  claim 2 , wherein the historic tax return data comprises one or more values that reflect an amount of tax listed due on a tax return in relation to an amount of tax listed as owed on the tax return. 
     
     
         10 . The method of  claim 2 , wherein the historic tax return data comprises one or move values that reflect a tax penalty amount listed on a tax return. 
     
     
         11 . The method of  claim 2 , wherein the historic tax return data comprises one or more values that reflect an amount of time after a tax deadline in which a tax return was filed. 
     
     
         12 . The method of  claim 2 , wherein the historic tax return data comprises one or more values that reflect an amount of tax that a tax return lists as owed. 
     
     
         13 . A computer-readable storage device encoded with a computer program, the program comprising instructions that, if executed by one or more computers, cause the one or more computers to perform operations comprising:
 generating training data for training a predictive model to estimate a likelihood of a debtor to pay a debt, the training data including, for each of a plurality of debtors, (i) historic tax return data for the debtor, (ii) commercial financial or credit data that is usable to calculate a credit score for the debtor, and (iii) a label indicating a likelihood of the debtor to pay a debt; and   training the predictive model using the historic tax return data, the commercial financial or credit data, and the labels included in the training data.   
     
     
         14 . The device of  claim 13 , wherein the historic tax return data comprises one or more values were entered on a personal income tax form by an associated debtor. 
     
     
         15 . The device of  claim 13 , wherein the debt comprises a debt owed to a government revenue agency. 
     
     
         16 . The device of  claim 13 , wherein the historic tax return data comprises data that is available only to a government revenue agency or to designated agents of the government revenue agency. 
     
     
         17 . The device of  claim 13 , wherein the historic tax return data comprises one or more values that reflect an amount of tax listed as due on a tax return in relation to an amount of income listed on the tax return. 
     
     
         18 . The device of  claim 13 , wherein the historic tax return data comprises one or more values that reflect an amount of tax listed due on a tax return in relation to an amount of tax listed as owed on the tax return. 
     
     
         19 . The device of  claim 13 , wherein the historic tax return data comprises one or move values that reflect a tax penalty amount listed on a tax return. 
     
     
         20 . The device of  claim 13 , wherein the historic tax return data comprises one or more values that reflect an amount of time after a tax deadline in which a tax return was filed. 
     
     
         21 . A system comprising:
 a processor configured to executed computer program instructions; and   a computer storage medium encoded with computer program instructions that, when executed by the processor, cause the system to perform operations comprising:   obtaining a predictive model that is trained to estimate a likelihood of a debtor to pay a debt, wherein the predictive model is trained using training data that includes, for each of a plurality of debtors, (i) historic tax return data for the debtor, (ii) commercial financial or credit data that is usable to calculate a credit score for the debtor, and (iii) a label indicating a likelihood of the debtor to pay a debt;   identifying a particular debtor;   obtaining (i) historic tax return data for the particular debtor, and (ii) commercial financial or credit data that is usable to calculate a credit score for the particular debtor;   providing, to the predictive model, (i) the historic tax return data for the particular debtor, and (ii) the commercial financial or credit data that is usable to calculate a credit score for the particular debtor; and
 obtaining, from the predictive model, an indication of a likelihood of the particular debtor to pay a debt.