Method and system of determining an optimal value of an auction parameter for a digital object

1. A method of determining an optimal value of an auction parameter for a digital object; the method executable at an auction server, associated with the storage; the auction server hosting an auction service, the method comprising:

acquiring from a storage, interaction history data of a first portion of users of the auction service, and interaction history data of a second portion of users of the auction service; the interaction history data including, for each respective interaction, at least: an indication of a digital object, an auction parameter associated with the digital object and an environment feature at a respective moment of time of the respective interaction;

based on the interaction history data associated with the first portion of users, executing an offline training of a machine learning algorithm to predict the optimal value of auction parameters for a plurality of digital objects, the plurality of digital objects being associated with the interaction history of the first portion of users;

applying the machine learning algorithm to determine a first optimal value of an auction parameter for a plurality of digital objects associated with the second portion of users;

storing the first optimal value of auction parameter for the plurality of digital objects;

responsive to receiving a request to the auction service by the auction server,

determining a digital object being relevant to the request;

determining a user associated with the digital object;

responsive to the user associated with the digital object being from the second portion of users, applying to the digital object the first optimal value of the auction parameter.

2. The method of claim 1 , further comprising setting a minimal number of users in the first portion of users of the auction service for the acquiring interaction history data.

3. The method of claim 1 , further comprising setting a minimal number of users in the second portion of users of the auction service for the acquiring interaction history data.

4. The method of claim 1 , wherein the number of users in the first portion of users of the auction service is equal to the number of users in the second portion.

5. The method of claim 1 , wherein during the acquiring interaction history data, the first and the second portions are presented as a single group of users of the auction service, and wherein prior to the executing the offline training of the machine learning algorithm, the single group of users is divided into the first portion of users of the auction service and the second portion of users of the auction service.

6. The method of claim 1 , further comprising:

prior to the executing the offline training of the machine learning algorithm, dividing a plurality of digital objects into at least two categories and dividing the first and the second portions of users of the auction service into subsets according to corresponding at least two categories; and wherein

the executing the offline training of the machine learning algorithm is based on the interaction history data associated with each distinct subset of the first portion of users;

applying the machine learning algorithm to determine a first optimal value of auction parameter for each category of a plurality of digital objects associated with a corresponding subset of the second portion of users;

upon receiving by the auction server a request to the auction service responsive to the user associated with the digital object being from the subset of the second portion of users, applying for the digital object the first optimal value of the auction parameter, corresponding to the digital object category.

7. The method of claim 6 , further comprising pre-setting a minimal number of users in each subset.

8. The method of claim 7 , wherein responsive to the number of users in a subset being less than the pre-set minimal value, executing the offline training of the machine learning algorithm is performed using the portion of users.

9. The method of claim 1 , wherein the storage includes a plurality of interaction history data for a plurality of users, and wherein the acquiring interaction history data comprises acquiring a sub-set of data generated at a pre-defined time period, the sub-set of data including for each respective interaction at least: the indication of the digital object, the auction parameter associated with the digital object and the environment feature at the respective moment of time of the respective interaction.

10. The method of claim 1 , wherein the interaction history data further includes a spatial position of the digital object at a screen of a computer device.

11. The method of claim 1 , wherein the interaction history data further includes an indication of a type of the digital object.

12. The method of claim 11 , wherein the type of the digital object is at least one of: a text, an image, a video, an animation, a button, a form, a hyperlink, an interactive element.

13. The method of claim 1 , wherein the interaction history data further includes a history of changes of at least one auction parameter over time.

14. The method of claim 1 , wherein feature of environment includes at least one of: an average bid, 90% bid quantile, a minimal bid, a maximal bid, a probability of a click on the digital object, a relevancy score of the digital object to a search query, an indication of a geographical area, and a search query, in response to which the digital object was shown.

15. The method of claim 1 , wherein the machine learning algorithm is configured to predict an optimal value of at least one of: a minimal placing price of the auction parameters, and an amnesty threshold for an advertiser.

16. The method of claim 1 , wherein after storing the first optimal value of the auction parameter for the plurality of the digital objects, the method further comprises updating the optimal value by periodical repetition of steps of the method of claim 1 .

17. The method of claim 1 , further comprising:

based on the interaction history data associated with the second portion of users, executing training of the machine learning algorithm to predict the optimal value of auction parameters for a plurality of digital objects, the plurality of digital objects being associated with the interaction history of the second portion of users;

applying the machine learning algorithm to determine a second optimal value of the auction parameter for the plurality of digital objects associated with the first portion of users;

storing the second optimal value of the auction parameter for the plurality of digital objects.

18. The method of claim 17 , wherein responsive to the user associated with the digital object being from the first portion of users, the method further comprises applying to the digital object the second optimal value of the auction parameter.

19. The method of claim 17 , wherein after storing the second optimal value of the auction parameter for the plurality of the digital objects, the method further comprises updating the optimal value by periodical repetition of steps of the method of claim 17 .