Methods and systems for determining advertising reach based on machine learning

Patent metadata

Field	Value
Publication number	US-2017193546-A1
Application number	US-201514985150-A
Country	US
Kind code	A1
Filing date	Dec 30, 2015
Priority date	Dec 30, 2015
Publication date	Jul 6, 2017
Grant date	—

Abstract

Methods and systems are provided for determining advertising reach based on machine learning. In particular, a reach calculator is provided to determine reach for advertisement campaigns in real time through the use of machine learning. The reach calculator increases the speed at which reach calculations can be done by using a trained machine learning model and a set of aggregated features as opposed to using a direct calculation approach that directly analyzes a massive amount of user data.

First claim

1 . A method for optimizing reach calculations, comprising:
retrieving a user data set; generating a set of aggregated features that is predictive of a reach of advertising campaigns, wherein the reach is a number of unique users who are exposed to an advertising campaign; developing a machine learning model by:
retrieving a sample user data set from the user data set based on a selected sample size;
determining a sample reach based on the set of aggregated features and the sample user data set;
determining, using the machine learning model, a simulated reach based on the set of aggregated features and the selected sample size;
determining whether a difference between the simulated reach and the sample reach exceeds a threshold; and
calibrating the machine learning model in response to determining that the difference exceeds the threshold, wherein the calibrating includes establishing a mathematical formula that defines a relationship between the simulated reach and the set of aggregated features; and
determining, on an on-demand basis, an estimate of the reach based on the set of aggregated features and the developed machine learning model.

2 . The method of claim 1 , further comprising:
retrieving a desired estimate of the reach; determining a difference between the determined estimate of the reach and the desired estimate of the reach; and in response to determining the difference, adjusting the advertising campaign, wherein the advertising campaign is adjustable at least by a number of advertisements included in the advertising campaign, advertisement frequencies, advertisement schedules, and advertisement channels.

3 . The method of claim 2 , further comprising:
determining, on an on-demand basis, a new estimate of the reach based on the set of aggregated features and the developed machine learning model after adjusting the advertising campaign; determining a new difference between the new determined estimate of the reach and the desired estimate of the reach; and in response to determining the difference, further adjusting the advertising campaign.

4 . The method of claim 2 , wherein the desired estimate of the reach is based on a user selection.

5 . The method of claim 1 , wherein the selected sample size is determined using a percentage of a total number of users.

6 . The method of claim 1 , wherein the calibrating the machine learning model comprises modifying a parameter of the machine learning model, and wherein the parameter is a variable that influences the relationship between the simulated reach and the set of aggregated features.

7 . The method of claim 1 , wherein the developing the machine learning model further comprises:
determining, using the machine learning model after the calibrating, a new simulated reach based on the set of aggregated features and the selected sample size; determining a new difference between the new simulated reach and the sample reach; and further calibrating the machine learning model in response to determining the new difference.

8 . The method of claim 1 , wherein the developing the machine learning model further comprises:
retrieving a new sample user data set from the user data set based on a new selected sample size; determining a new sample reach based on the set of aggregated features and the new sample user data set; determining, using the machine learning model after the calibrating, a new simulated reach based on the set of aggregated features and the new selected sample size; determining a new difference between the new simulated reach and the new sample reach; and further calibrating the machine learning model in response to determining the new difference.

9 . The method of claim 1 , wherein the set of aggregated features is based on a user selection.

10 . The method of claim 1 , wherein the set of aggregated features is based on a machine selection.

11 . A system for optimizing reach calculations, the system comprising:
control circuitry configured to:
retrieve a user data set;
generate a set of aggregated features that is predictive of a reach of advertising campaigns, wherein the reach is a number of unique users who are exposed to an advertising campaign;
develop a machine learning model by:
retrieving a sample user data set from the user data set based on a selected sample size;
determining a sample reach based on the set of aggregated features and the sample user data set;
determining, using the machine learning model, a simulated reach based on the set of aggregated features and the selected sample size;
determining whether a difference between the simulated reach and the sample reach exceeds a threshold; and
calibrating the machine learning model in response to determining that the difference exceeds the threshold, wherein the calibrating includes establishing a mathematical formula that defines a relationship between the simulated reach and the set of aggregated features; and

determine, on an on-demand basis, an estimate of the reach based on the set of aggregated features and the developed machine learning model.

12 . The system of claim 11 , wherein the control circuitry is further configured to:
retrieve a desired estimate of the reach; determine a difference between the determined estimate of the reach and the desired estimate of the reach; and in response to determining the difference, adjust the advertising campaign, wherein the advertising campaign is adjustable at least by a number of advertisements included in the advertising campaign, advertisement frequencies, advertisement schedules, and advertisement channels.

13 . The system of claim 12 , wherein the control circuitry is further configured to:
determine, on an on-demand basis, a new estimate of the reach based on the set of aggregated features and the developed machine learning model after adjusting the advertising campaign; determine a new difference between the new determined estimate of the reach and the desired estimate of the reach; and in response to determining the difference, further adjust the advertising campaign.

14 . The method of claim 12 , wherein the desired estimate of the reach is based on a user selection.

15 . The system of claim 11 , wherein the selected sample size is determined using a percentage of a total number of users.

16 . The system of claim 11 , wherein the control circuitry configured to calibrate the machine learning model is further configured to modify a parameter of the machine learning model, and wherein the parameter is a variable that influences the relationship between the simulated reach and the set of aggregated features.

17 . The system of claim 11 , wherein the control circuitry configured to develop the machine learning model is further configured to:
determine, using the machine learning model after the calibrating, a new simulated reach based on the set of aggregated features and the selected sample size; determine a new difference between the new simulated reach and the sample reach; and further calibrate the machine learning model in response to determining the new difference.

18 . The system of claim 11 , wherein the control circuitry configured to develop the machine learning model is further configured to:
retrieve a new sample user data set from the user data set based on a new selected sample size; determine a new sample reach based on the set of aggregated features and the new sample user data set; determine, using the machine learning model after the calibrating, a new simulated reac

Assignees

• Rovi Guides Inc

Inventors

• Bennett Steven
• Kelley Randall
• Xu Xiaoxi

Classifications

• G06Q30/0244 Primary

  Optimization · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q30/0201

  Market modelling; Market analysis; Collecting market data · CPC title

• G06N99/005

  Physics · mapped topic