System, Method, and Computer Program Product for Determining a Control Account that Corresponds to an Exposed Account

A computer program product, the computer program product comprising at least one non-transitory computer-readable medium including one or more instructions that, when executed by at least one processor, cause the at least one processor to:

• determine a first cohort level group of accounts, wherein the first cohort level group of accounts comprises a group of exposed accounts and a group of control accounts that are associated with a first specified event time period of a plurality of event time periods, wherein each account of the group of exposed accounts is associated with an event time period of a plurality of event time periods, wherein each event time period of the plurality of event time periods is associated with a time period during which the account was exposed to an event associated with a merchant, wherein the event associated with the merchant is an offer associated with a loyalty program of the merchant or enrollment in the loyalty program of the merchant, and wherein each account of the group of control accounts conducted at least one transaction involving the merchant during at least one event time period of the plurality of event time periods;
• determine a first segment level group of accounts from the first cohort level group of accounts;
• generate a prediction model based on a plurality of control accounts that are included in the first segment level group of accounts, wherein the prediction model is configured to output a prediction score that represents an amount of spending involving an account after the first specified event time period, wherein, when generating the prediction model, the at least one processor is programmed or configured to:
• determine a plurality of modeling variables for each account of the group of control accounts included in the first cohort level group of accounts, wherein the plurality of modeling variables comprises:
• a number of transactions initiated using the account that involve the merchant during the first specified event time period;
• determine a variable importance metric associated with each modeling variable of the plurality of modeling variables;
• exclude one or more modeling variables of the plurality of modeling variables based on the variable importance metric associated with the one or more modeling variables of the plurality of modeling variables to provide a subset of modeling variables;
• train the prediction model based on the subset of modeling variables;
• calculate an error term based on training the prediction model;
• and
• update one or more weights of the prediction model based on the error term;
• identify a first exposed account of a plurality of exposed accounts that are included in the first segment level group of accounts;
• and
• determine a first control account of the plurality of control accounts that are included in the first segment level group of accounts that corresponds to the first exposed account using the prediction model.

A method, comprising:

• determining, with at least one processor, a first cohort level group of accounts, wherein the first cohort level group of accounts comprises a group of exposed accounts and a group of control accounts that are associated with a first specified event time period of a plurality of event time periods, wherein each account of the group of exposed accounts is associated with an event time period of a plurality of event time periods, wherein each event time period of the plurality of event time periods is associated with a time period during which the account was exposed to an event associated with a merchant, wherein the event associated with the merchant is an offer associated with a loyalty program of the merchant or enrollment in the loyalty program of the merchant, and wherein each account of the group of control accounts conducted at least one transaction involving the merchant during at least one event time period of the plurality of event time periods;
• determining, with at least one processor, a first segment level group of accounts from the first cohort level group of accounts;
• generating, with at least one processor, a prediction model based on a plurality of control accounts that are included in the first segment level group of accounts, wherein the prediction model is configured to output a prediction score that represents an amount of spending involving an account after the first specified event time period, wherein generating the prediction model comprises:
• determining a plurality of modeling variables for each account of the group of control accounts included in the first cohort level group of accounts, wherein the plurality of modeling variables comprises:
• a number of transactions initiated using the account that involve the merchant during the first specified event time period;
• determining a variable importance metric associated with each modeling variable of the plurality of modeling variables;
• excluding one or more modeling variables of the plurality of modeling variables based on the variable importance metric associated with the one or more modeling variables of the plurality of modeling variables to provide a subset of modeling variables;
• training the prediction model based on the subset of modeling variables;
• calculating an error term based on training the prediction model;
• and
• updating one or more weights of the prediction model based on the error term;
• identifying, with at least one processor, a first exposed account of a plurality of exposed accounts that are included in the first segment level group of accounts;
• and
• determining, with at least one processor, a first control account of the plurality of control accounts that are included in the first segment level group of accounts that corresponds to the first exposed account using the prediction model.