US 12,229,798 B2
System and method for individualized exposure estimation in linear media advertising for cross platform audience management and other applications
Bradley Imbach, Roswell, GA (US); Wassim Chaar, Coppell, TX (US); Howard Shimmel, Wilton, CT (US); and Daniel Aversano, Huntington, NY (US)
Assigned to datafuelX Inc., Chappaqua, NY (US)
Filed by datafuelX Inc., Chappaqua, NY (US)
Filed on Oct. 5, 2022, as Appl. No. 17/938,127.
Claims priority of provisional application 63/253,331, filed on Oct. 7, 2021.
Prior Publication US 2023/0110511 A1, Apr. 13, 2023
Int. Cl. G06Q 30/00 (2023.01); G06Q 30/0242 (2023.01)
CPC G06Q 30/0244 (2013.01) 19 Claims
OG exemplary drawing
 
1. A method for forecasting advertisement exposure, the method comprising:
receiving, with a processor, information detailing a plurality of future linear media programs and information detailing a plurality of advertisement spots to be aired during the plurality of future linear media programs;
determining, with the processor, for each respective viewer of a plurality of viewers, a respective plurality of predicted proportions of the plurality of future linear media programs that will be viewed by the respective viewer, using at least one machine learning model and based on the information detailing the plurality of future linear media programs;
determining, with the processor, for each respective viewer of the plurality of viewers, whether the respective viewer will be exposed to at least one of the plurality of advertisement spots based on (i) the respective plurality of predicted proportions and (ii) the information detailing the plurality of advertisement spots; and
training the at least one machine learning model based on historical viewing activity data for the plurality of viewers, the historical viewing activity data including (i) information detailing a plurality of historical linear media programs and (ii) a plurality of observed proportions of the plurality of historical linear media programs that were viewed by the plurality of viewers, parameters of the at least one machine learning model being optimized in an iterative manner based on errors between the plurality of observed proportions and outputs of the at least one machine learning model.