Performing predictive inferences using multiple predictive models

Abstract

Method, apparatus and computer program product for performing a cross-model predictive inference to generate a cross-model predictive output for a plurality of predictive inputs using a plurality of predictive models. For example, the apparatus includes at least one processor and at least one non-transitory memory including program code. The at least one non-transitory memory and the program code are configured to, with the at least one processor, obtain a model selection probability distribution which defines, for each predictive model, a respective selection probability score; obtain, for each predictive model, respective cross-model normalization data; for each predictive input, determine a cross-model predictive score; and determine, based on each determined cross-model predictive score, the cross-model predictive output.

1 - 22 . (canceled) 
     
     
         23 . An apparatus comprising at least one processor and at least one non-transitory memory storing instructions that, with the at least one processor, cause the apparatus to:
 determine a cross-model predictive score for each predictive input of a plurality of predictive inputs by:
 determining, based at least in part on a weighted random selection for the predictive input, a respective selected predictive model of a plurality of predictive models, wherein at least one weighted random selection parameter is determined based at least in part on respective selection probability scores of the plurality of predictive models; 
 generating, based at least in part on one or more predictive operations of the respective selected predictive model and the predictive input, one or more model-specific predictive scores for the predictive input; and 
 determining the cross-model predictive score for the predictive input based at least in part on transforming the one or more model-specific predictive scores associated with the predictive input using respective cross-model normalization parameters of cross-model normalization data for the selected predictive model associated with the predictive input; 
   generate, based at least in part on one or more optimal predictive models and one or more predictive inputs of the predictive ranking of predictive inputs, cross-model predictive output;   generate, based at least in part on the cross-model predictive output, an electronic communication; and   transmit the electronic communication to a computing device, wherein the electronic communication is configured to render data via a graphical interface of the computing device.   
     
     
         24 . The apparatus of  claim 23 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 obtain a model selection probability distribution, wherein the model selection probability distribution defines, for each predictive model of the plurality of predictive models, the respective selection probability score.   
     
     
         25 . The apparatus of  claim 23 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 obtain, for each predictive model of the plurality of predictive models, the respective cross-model normalization data.   
     
     
         26 . The apparatus of  claim 23 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 determine, based at least in part on a predictive ranking of predictive inputs of the plurality of predictive inputs ranked according to respective cross-model predictive scores for the predictive inputs, the one or more optimal predictive models of the plurality of predictive models.   
     
     
         27 . The apparatus of  claim 23 , wherein the electronic communication is associated with a merchant of goods or services. 
     
     
         28 . The apparatus of  claim 23 , wherein the weighted random selection for a predictive input of the plurality of predictive inputs is characterized by one or more weighted random selection parameters for the predictive input. 
     
     
         29 . The apparatus of  claim 23 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 identify, from the plurality of predictive models, a champion predictive model and one or more challenger predictive models.   
     
     
         30 . A computer-implemented method, comprising:
 determining a cross-model predictive score for each predictive input of a plurality of predictive inputs by:
 determining, based at least in part on a weighted random selection for the predictive input, a respective selected predictive model of a plurality of predictive models, wherein at least one weighted random selection parameter is determined based at least in part on respective selection probability scores of the plurality of predictive models; 
 generating, based at least in part on one or more predictive operations of the respective selected predictive model and the predictive input, one or more model-specific predictive scores for the predictive input; and 
 determining the cross-model predictive score for the predictive input based at least in part on transforming the one or more model-specific predictive scores associated with the predictive input using respective cross-model normalization parameters of cross-model normalization data for the selected predictive model associated with the predictive input; 
   generating, based at least in part on one or more optimal predictive models and one or more predictive inputs of the predictive ranking of predictive inputs, cross-model predictive output;   generating, based at least in part on the cross-model predictive output, an electronic communication; and   transmitting the electronic communication to a computing device, wherein the electronic communication is configured to render data via a graphical interface of the computing device.   
     
     
         31 . The method of  claim 30 , further comprising:
 obtaining a model selection probability distribution, wherein the model selection probability distribution defines, for each predictive model of the plurality of predictive models, the respective selection probability score.   
     
     
         32 . The method of  claim 30 , further comprising:
 obtaining, for each predictive model of the plurality of predictive models, the respective cross-model normalization data.   
     
     
         33 . The method of  claim 30 , further comprising:
 determining, based at least in part on a predictive ranking of predictive inputs of the plurality of predictive inputs ranked according to respective cross-model predictive scores for the predictive inputs, the one or more optimal predictive models of the plurality of predictive models.   
     
     
         34 . The method of  claim 30 , wherein the electronic communication is associated with a merchant of goods or services. 
     
     
         35 . The method of  claim 30 , wherein the weighted random selection for a predictive input of the plurality of predictive inputs is characterized by one or more weighted random selection parameters for the predictive input. 
     
     
         36 . The method of  claim 30 , further comprising:
 identifying, from the plurality of predictive models, a champion predictive model and one or more challenger predictive models.   
     
     
         37 . At least one non-transitory memory storing instructions that, with at least one processor, cause an apparatus to:
 determine a cross-model predictive score for each predictive input of a plurality of predictive inputs by:
 determining, based at least in part on a weighted random selection for the predictive input, a respective selected predictive model of a plurality of predictive models, wherein at least one weighted random selection parameter is determined based at least in part on respective selection probability scores of the plurality of predictive models; 
 generating, based at least in part on one or more predictive operations of the respective selected predictive model and the predictive input, one or more model-specific predictive scores for the predictive input; and 
 determining the cross-model predictive score for the predictive input based at least in part on transforming the one or more model-specific predictive scores associated with the predictive input using respective cross-model normalization parameters of cross-model normalization data for the selected predictive model associated with the predictive input; 
   generate, based at least in part on one or more optimal predictive models and one or more predictive inputs of the predictive ranking of predictive inputs, cross-model predictive output;   generate, based at least in part on the cross-model predictive output, an electronic communication; and   transmit the electronic communication to a computing device, wherein the electronic communication is configured to render data via a graphical interface of the computing device.   
     
     
         38 . The at least one non-transitory memory of  claim 37 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 obtain, for each predictive model of the plurality of predictive models, the respective cross-model normalization data.   
     
     
         39 . The at least one non-transitory memory of  claim 37 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 determine, based at least in part on a predictive ranking of predictive inputs of the plurality of predictive inputs ranked according to respective cross-model predictive scores for the predictive inputs, the one or more optimal predictive models of the plurality of predictive models.   
     
     
         40 . The at least one non-transitory memory of  claim 37 , wherein the electronic communication is associated with a merchant of goods or services. 
     
     
         41 . The at least one non-transitory memory of  claim 37 , wherein the weighted random selection for a predictive input of the plurality of predictive inputs is characterized by one or more weighted random selection parameters for the predictive input. 
     
     
         42 . The at least one non-transitory memory of  claim 37 , wherein the at least one non-transitory memory stores instructions that, with the at least one processor, further cause the apparatus to:
 identify, from the plurality of predictive models, a champion predictive model and one or more challenger predictive models.