System and method for building a predictive score without model training

Abstract

A system and method for building a predictive score without model training are disclosed. A set of predictive variables is defined based on raw data fields generated from raw data from one or more sources and domain knowledge. The raw data includes a historical set of transactions previously generated by one or more raw data sources. An scaled relative risk table to describe each predictive variable of the set of predictive variables is generated. The set of predictive variables is combined based on their associated relative risk tables to generate a predictive score for a future set of transactions.

What is claimed is: 
     
         1 . A computer-implemented method of building a predictive score without model training, the method comprising:
 defining, by a computer, a set of predictive variables based on raw data fields generated from raw data from one or more sources, the raw data including a historical set of transactions previously generated by one or more raw data sources;   generating, by the computer, a relative risk table to describe each predictive variable of the set of predictive variables;   adapting each predictive variable to an average value of one;   assigning, by the computer, an adapted relative risk to each predictive variable;   combining, by the computer, the set of predictive variables having the average value of one using their associated relative risk tables; and   generating a predictive score for a future set of transactions based on the combined set of predictive variables and their assigned relative risks.   
     
     
         2 . The method in accordance with  claim 1 , wherein the raw data includes domain knowledge of the set of transactions. 
     
     
         3 . A computer-implemented method of building a predictive score without model training, the method comprising:
 accessing, by a computer, raw data from one or more sources, the raw data including a historical set of transactions previously generated by one or more raw data sources;   defining raw data fields from the raw data;   defining a set of predictive variables based on the raw data fields generated from raw data from one or more raw data sources;   generating, by the computer, a relative risk table to describe each predictive variable of the set of predictive variables;   adapting each predictive variable to an average value of one;   assigning an adapted relative risk to each predictive variable;   combining the set of predictive variables having the average value of one using their associated relative risk tables; and   generating, by the computer, a predictive score for a future set of transactions according to the combined set of predictive variables and their associated assigned relative risks.   
     
     
         4 . The method in accordance with  claim 3 , wherein the raw data includes domain knowledge of the set of transactions. 
     
     
         5 . A system for building a predictive score without model training, the system comprising:
 a computing system including a processor for executing instructions encoded in a tangible medium, the instructions comprising:
 a data fields definition module for defining a set of predictive variables based on raw data fields generated from raw data from one or more raw data sources; 
   a relative risk table generation module for generating a relative risk table to describe each predictive variable of the set of predictive variables, and for assigning an adapted relative risk to each predictive variable;
 an average conversion module for adapting each predictive variable to an average value of one; 
 a variable combination and score generation module for combining the set of predictive variables having the average value of one using their associated relative risk tables, and for generating a predictive score for a future set of transactions. 
   
     
     
         6 . The system in accordance with  claim 5 , wherein the raw data includes domain knowledge of the set of transactions. 
     
     
         7 . The system in accordance with  claim 5 , further comprising a communications network connected between the computing system and the one or more raw data sources. 
     
     
         8 . The system in accordance with  claim 7 , wherein the communications network includes an extranet.