Methods of processing and segmenting web usage information
First Claim
Patent Images
1. A method comprising:
- creating, by a processor, a baseline behavioral targeting system using a machine learning system comprising;
retrieving a training data set comprising page view data, ad view data, and ad click data, andtraining the machine learning system using the training data set;
calculating, by the processor, a baseline performance metric value for said baseline behavioral targeting system a test data set comprising;
retrieving existing source data and existing result data, the existing source data comprising webpage viewing habits and demographic information of users, the existing result data comprising information for indicating whether a user has clicked on advertisements on at least one webpage that were selected by using the existing source data and an amount of time spent at an associated advertiser'"'"'s web site after a click;
inputting the existing source data into the baseline behavioral targeting system,calculating predicted result data using the baseline behavioral targeting system, andcomparing the predicted result data to the existing result data to calculate the baseline performance metric value;
creating, by the processor, a proposed behavioral targeting system using the machine learning system comprising;
retrieving a proposed training data set comprising the training data set and new search data, andtraining the machine learning system using the proposed training data set;
calculating, by the processor, a second performance metric value for said proposed behavioral targeting system comprising;
retrieving the existing source data and the existing result data;
inputting the existing source data into the proposed behavioral targeting system,calculating proposed predicted result data using the proposed behavioral targeting system, andcomparing the proposed predicted result data to the existing result data to calculate the second performance metric value;
determining, by the processor, a value of the new search data by comparing said second performance metric value to said baseline performance metric value, the value indicating whether the new search data improved the predicted result data of the proposed behavioral targeting system in comparison with the predicted result data of the baseline behavioral targeting system;
when said second performance metric value is greater than said baseline performance metric value, using said new search data for selecting advertisements for display;
when said baseline performance metric value is greater than said second performance metric value, using said existing source data for selecting advertisements for display; and
displaying, by the processor, the selected advertisements on a webpage.
9 Assignments
0 Petitions
Accused Products
Abstract
A current behavioral targeting system is first tested using a suite of test data. The output of the test is one or more performance metrics. Next, newly proposed behavioral targeting system created. The newly proposed behavioral targeting system is then evaluated using both the existing source data and a new source data. The evaluation of the newly proposed behavioral targeting system produces one or more performance metrics of the same type earlier calculated. Finally, the two sets of performance metrics are compared. The performance metric difference represents the impact of the new source data.
12 Citations
16 Claims
-
1. A method comprising:
-
creating, by a processor, a baseline behavioral targeting system using a machine learning system comprising; retrieving a training data set comprising page view data, ad view data, and ad click data, and training the machine learning system using the training data set; calculating, by the processor, a baseline performance metric value for said baseline behavioral targeting system a test data set comprising; retrieving existing source data and existing result data, the existing source data comprising webpage viewing habits and demographic information of users, the existing result data comprising information for indicating whether a user has clicked on advertisements on at least one webpage that were selected by using the existing source data and an amount of time spent at an associated advertiser'"'"'s web site after a click; inputting the existing source data into the baseline behavioral targeting system, calculating predicted result data using the baseline behavioral targeting system, and comparing the predicted result data to the existing result data to calculate the baseline performance metric value; creating, by the processor, a proposed behavioral targeting system using the machine learning system comprising; retrieving a proposed training data set comprising the training data set and new search data, and training the machine learning system using the proposed training data set; calculating, by the processor, a second performance metric value for said proposed behavioral targeting system comprising; retrieving the existing source data and the existing result data; inputting the existing source data into the proposed behavioral targeting system, calculating proposed predicted result data using the proposed behavioral targeting system, and comparing the proposed predicted result data to the existing result data to calculate the second performance metric value; determining, by the processor, a value of the new search data by comparing said second performance metric value to said baseline performance metric value, the value indicating whether the new search data improved the predicted result data of the proposed behavioral targeting system in comparison with the predicted result data of the baseline behavioral targeting system; when said second performance metric value is greater than said baseline performance metric value, using said new search data for selecting advertisements for display; when said baseline performance metric value is greater than said second performance metric value, using said existing source data for selecting advertisements for display; and displaying, by the processor, the selected advertisements on a webpage. - View Dependent Claims (2, 3, 4, 5)
-
-
6. A non-transitory computer-readable storage medium, said computer-readable storage medium comprising a set of instructions for evaluating behavioral targeting source data, said set of instructions implementing the steps of:
-
creating a baseline behavioral targeting system using a machine learning system comprising; retrieving a training data set comprising page view data, ad view data, and ad click data, and training the machine learning system using the training data set; calculating a baseline performance metric value for said baseline behavioral targeting system comprising; retrieving existing source data and existing result data, the existing source data comprising webpage viewing habits and demographic information of users, the existing result data comprising information for indicating whether a user has clicked on advertisements on at least one webpage that were selected by using the existing source data and an amount of time spent at an associated advertiser'"'"'s web site after a click; inputting the existing source data into the baseline behavioral targeting system, calculating predicted result data using the baseline behavioral targeting system, and comparing the predicted result data to the existing result data to calculate the baseline performance metric value; creating a proposed behavioral targeting system using the machine learning system comprising; retrieving a proposed training data set comprising the training data set and new search data, and training the machine learning system using the proposed training data set; calculating a second performance metric value for said proposed behavioral targeting system comprising; retrieving the existing source data and the existing result data; inputting the existing source data into the proposed behavioral targeting system, calculating proposed predicted result data using the proposed behavioral targeting system, and comparing the proposed predicted result data to the existing result data to calculate the second performance metric value; determining a value of the new search data by comparing said second performance metric value to said baseline performance metric value, the value indicating whether the new search data improved the predicted result data of the proposed behavioral targeting system in comparison with the predicted result data of the baseline behavioral targeting system; when said second performance metric value is greater than said baseline performance metric value, using said new search data to select a first new advertisement for display; when said baseline performance metric value is greater than said second performance metric value, using said existing source data to select a second new advertisement for display; displaying, by the processor, the selected advertisements on a webpage. - View Dependent Claims (7, 8, 9, 10, 11)
-
-
12. A method comprising:
-
creating, by a processor, a baseline behavioral targeting system using a machine learning system comprising; retrieving a training data set comprising page view data, ad view data, and ad click data, and training the machine learning system using the training data set; calculating, by a processor, a baseline performance metric value for said baseline behavioral targeting system comprising; retrieving existing source data and existing result data, the existing source data comprising webpage viewing habits and demographic information of users, the existing result data comprising information for indicating whether a user has clicked on advertisements on at least one webpage that were selected by using the existing source data and an amount of time spent at an associated advertiser'"'"'s web site after a click; inputting the existing source data into the baseline behavioral targeting system, calculating predicted result data using the baseline behavioral targeting system, and comparing the predicted result data to the existing result data to calculate the baseline performance metric value; creating, by the processor, a proposed behavioral targeting system using the machine learning system comprising; retrieving a proposed training data set comprising the training data set and new source data, the new source data comprising additional webpage viewing habits of users, geographic location information, and additional demographic information of users, and training the machine learning system using the proposed training data set; calculating, by the processor, a second performance metric value for said proposed behavioral targeting system comprising; retrieving the existing source data and the existing result data; inputting the source data into the proposed behavioral targeting system, calculating proposed predicted result data using the proposed behavioral targeting system, and comparing the proposed predicted result data to the existing result data to calculate the second performance metric value; determining, by the processor, a value of said new source data by comparing said second performance metric value to said baseline performance metric value, the determination indicating whether the new source data improved the predicted result data of the proposed behavioral targeting system in comparison with the predicted result data of the baseline behavioral targeting system; when said second performance metric value is greater than said baseline performance metric value, using said new source data to select a first new advertisement for display; when said baseline performance metric value is greater than said second performance metric value, using said existing source data to select a second new advertisement for display; displaying, by the processor, the selected advertisements on the at least one webpage. - View Dependent Claims (13, 14, 15, 16)
-
Specification