User advertisement click behavior modeling

1. A method performed on a computing device, the method for determining click fraud, the method comprising:

accessing log data that represents search events and corresponding click events, where the log data is based on a plurality of user devices performing the search events and the corresponding click events, where each of the search events occurs when a corresponding one of the plurality of user devices performs a query via a search internet protocol (“IP”) address, and where each of the corresponding click events occurs when the corresponding one of the plurality of user devices selects, via a click IP address, an advertisement included in a search result provided in response to the performed query, where, for each of the search events, the log data comprises a corresponding search time that indicates a time that the corresponding search result was generated, the corresponding query, and a request globally unique identifier (“RGUID”) that uniquely identifies the corresponding performed query and that is assigned to the corresponding search result provided in response to the corresponding performed query, and where, for each of the click events, the log data further comprises a corresponding click time that indicates a time that the corresponding advertisement was selected;

constructing a user linkage representation of the accessed log data based on a set time period that is based on a portion of the search times and the click times, where the set time period covers the portion of the search times and the click times, and where, for each of the search events and the corresponding click events within the set time period, the search IP address is linked to the click IP address in the user linkage representation;

mining similar user clusters from the user linkage representation, where the mined similar user clusters share search and click patterns identified in the user linkage representation; and

identifying, based on the shared search and click patterns, ones of the plurality of user devices associated with the similar user clusters as candidate devices for committing click fraud.

2. The method as recited in claim 1 , wherein the log data comprises search log data and click log data.

3. The method as recited in claim 2 , wherein the search log data represents a search event triggered by the query, and where the click log data describes a click event associated with the search event.

4. The method as recited in claim 3 , wherein the results of the query are identified by a search globally unique identifier.

5. The method as recited in claim 1 further comprising, analyzing the user linkage representation with respect to the search time and the click time.

6. The method as recited in claim 5 , wherein the identifying the respective ones of the user devices further comprises classifying the respective ones of the user devices as strong suspect devices for committing click fraud based on at least two of the similar user clusters.

7. At least one computer memory device storing instructions that, when executed by a computing device, cause the computing device to perform actions for determining click fraud, the actions comprising:

accessing log data that represents search events and corresponding click events, where the log data is based on a plurality of user devices performing the search events and the corresponding click events, where each of the search events occurs when a corresponding one of the plurality of user devices performs a query via a search internet protocol (“IP”) address, and where each of the corresponding click events occurs when the corresponding one of the plurality of user devices selects, via a click IP address, an advertisement included in a search result provided in response to the performed query, where, for each of the search events, the log data comprises a corresponding search time that indicates a time that the corresponding search result was generated, the corresponding query, and a request globally unique identifier (“RGUID”) that uniquely identifies the corresponding performed query and that is assigned to the corresponding search result provided in response to the corresponding performed query, and where, for each of the click events, the log data further comprises a corresponding click time that indicates a time that the corresponding advertisement was selected;

constructing a user linkage representation of the accessed log data based on a set time period that is based on a portion of the search times and the click times, where the set time period covers the portion of the search times and the click times, and where, for each of the search events and the corresponding click events within the set time period, the search IP address is linked to the click IP address in the user linkage representation;

mining similar user clusters from the user linkage representation, where the mined similar user clusters share search and click patterns identified in the user linkage representation; and

identifying, based on the shared search and click patterns, ones of the plurality of user devices associated with the similar user clusters as candidate devices for committing click fraud.

8. The at least one computer memory device as recited in claim 7 , where the log data comprises search log data and click log data.

9. The at least one computer memory device as recited in claim 8 , where the search log data represents a search event triggered by the query, and where the click log data describes a click event associated with the search event.

10. The at least one computer memory device as recited in claim 9 , where the results of the query are identified by a search globally unique identifier.

11. A system comprising a computer and software together configured for performing actions for determining click fraud, the actions comprising:

accessing log data that represents search events and corresponding click events, where the log data is based on a plurality of user devices performing the search events and the corresponding click events, where each of the search events occurs when a corresponding one of the plurality of user devices performs a query via a search internet protocol (“IP”) address, and where each of the corresponding click events occurs when the corresponding one of the plurality of user devices selects, via a click IP address, an advertisement included in a search result provided in response to the performed query, where, for each of the search events, the log data comprises a corresponding search time that indicates a time that the corresponding search result was generated, the corresponding query, and a request globally unique identifier (“RGUID”) that uniquely identifies the corresponding performed query and that is assigned to the corresponding search result provided in response to the corresponding performed query, and where, for each of the click events, the log data further comprises a corresponding click time that indicates a time that the corresponding advertisement was selected;

constructing, by the computer, a user linkage representation of the accessed log data based on a set time period that is based on a portion of the search times and the click times, where the set time period covers the portion of the search times and the click times, and where, for each of the search events and the corresponding click events within the set time period, the search IP address is linked to the click IP address in the user linkage representation;

mining similar user clusters from the user linkage representation, where the mined similar user clusters share search and click patterns identified in; and

identifying, based on the shared search and click patterns, ones of the plurality of user devices associated with the similar user clusters as candidate devices for committing click fraud.

12. The system of claim 11 where the log data comprises search log data and click log data.

13. The system of claim 12 where the search log data represents a search event device triggered by the query, and where the click log data describes a click event associated with the search event.

14. The system of claim 13 where the results of the query are identified by a search globally unique identifier.

15. The system of claim 11 , the actions further comprising analyzing the user linkage representation with respect to the search time and the click time.