Systems and methods for user propensity classification and online auction design

1. A server computing device, comprising:

a network interface configured to:

receive a detection message from a publisher server, the detection message including user activity information on a webpage operated by the publisher server;

receive a bid request from an advertiser server; and

receive dimensional information; and

a processor in communication with the network interface, the processor configured to:

generate a sample based on at least one of the detection message, the bid request, or the dimensional information;

run the sample through a set of cascading classifier models, wherein the running the sample through the set of cascading classifier models comprises:

inputting the sample to a first classifier model of the set of cascading classifier models;

determining whether the sample is associated with a value exceeding a predetermined revenue threshold of the first classifier model; and

responsive to determining that the sample is associated with a value exceeding the predetermined revenue threshold, inputting the sample to a second classifier model of the set of cascading classifier models;

predict a floor price value based on the sample running through the set of cascading classifier models; and

implement an online bid auction based on the floor price value.

2. The server computing device of claim 1 , wherein each subsequent classifier model in the set of cascading classifier models is more restricting than a prior.

3. The server computing device of claim 1 , wherein each classifier model in the set of cascading classifier models is trained by machine learning according to an Adaboost technique.

4. The server computing device of claim 1 , wherein the processor is configured to:

implement a second price online bid auction that incorporates the floor price value; and

select the bid request as a winning bid of the second price online bid auction when the bid request has a value greater than other received bid request values and greater than the floor price value.

5. The server computing device of claim 1 , wherein the bid request is allocated to a price range bucket from a plurality of price range buckets, wherein the price range bucket represents a range of floor price values.

6. The server computing device of claim 1 , wherein the implementing the online bid auction comprises:

applying the floor price value to a winning bid in the online bid auction.

7. The server computing device of claim 1 , wherein the processor is configured to:

determine whether a spread between a first received bid and a second received bid is greater than a predetermined bid spread; and

when the spread between the first received bid and the second received bid is less than the predetermined bid spread, withhold running the sample through the set of cascading classifier models.

8. The server computing device of claim 1 , wherein the dimensional information comprises at least one of a search query information, a communication device information, a publisher information, a historical online activity information, or a user attribute information.

9. A method for predicting a floor price, comprising:

receiving, by a processor, a detection message sent from a publisher server, the detection message including user activity information on a webpage operated by the publisher server;

receiving, by the processor, a bid request sent from an advertiser server;

receiving, by the processor, dimensional information;

generating, by the processor, a sample based on at least one of the detection message, the bid request, or the dimensional information;

running, by the processor, the sample through a set of cascading classifier models, wherein the running the sample through the set of cascading classifier models comprises:

inputting the sample to a first classifier model of the set of cascading classifier models;

determining whether the sample is associated with a value exceeding a predetermined revenue threshold of the first classifier model; and

responsive to determining that the sample is associated with a value exceeding the predetermined revenue threshold, inputting the sample to a second classifier model of the set of cascading classifier models;

predicting, by the processor, a floor price value based on the sample running through the set of cascading classifier models; and

implementing an online bid auction based on the floor price value.

10. The method of claim 9 , wherein each subsequent classifier model in the set of cascading classifier models is more restricting than a prior.

11. The method of claim 9 , wherein each classifier model in the set of cascading classifier models is trained by machine learning according to an Adaboost technique.

12. The method of claim 9 , comprising:

implementing, by the processor, a second price online bid auction that incorporates the floor price value; and

selecting, by the processor, the bid request as a winning bid of the second price online bid auction when the bid request has a value greater than other received bid request values and greater than the floor price value.

13. The method of claim 9 , wherein the bid request is allocated to a price range bucket from a plurality of price range buckets, wherein the price range bucket represents a range of floor price values.

14. The method of claim 9 , wherein the implementing the online bid auction comprises:

applying the floor price value to a winning bid in the online bid auction.

15. The method of claim 9 , comprising:

determining, by the processor, whether a spread between a first received bid and a second received bid is greater than a predetermined bid spread; and

when the spread between the first received bid and the second received bid is less than the predetermined bid spread, withholding, by the processor, running the sample through the set of cascading classifier models.

16. A non-transitory computer readable medium storing a set of processor executable instructions that, when executed by a processor, cause the processor to:

receive a detection message sent from a publisher server, the detection message including user activity information on a webpage operated by the publisher server;

receive a bid request sent from an advertiser server;

receive dimensional information;

generate a sample based on at least one of the detection message, the bid request, or the dimensional information;

run the sample through a set of cascading classifier models, wherein the running the sample through the set of cascading classifier models comprises:

inputting the sample to a first classifier model of the set of cascading classifier models;

determining whether the sample is associated with a value exceeding a threshold of the first classifier model; and

responsive to determining that the sample is associated with a value exceeding the threshold, inputting the sample to a second classifier model of the set of cascading classifier models;

predict a floor price value based on the sample running through the set of cascading classifier models; and

implement an online bid auction based on the floor price value.

17. The non-transitory computer readable medium of claim 16 , storing instructions that, when executed by the processor, cause the processor to:

implement a second price online bid auction that incorporates the floor price value; and

select the bid request as a winning bid of the second price online bid auction when the bid request has a value greater than other received bid request values and greater than the floor price value.

18. The non-transitory computer readable medium of claim 16 , wherein the implementing the online bid auction comprises:

applying the floor price value to a winning bid in the online bid auction.

19. The non-transitory computer readable medium of claim 16 , storing instructions that, when executed by the processor, cause the processor to:

determine whether a spread between a first received bid and a second received bid is greater than a predetermined bid spread; and

when the spread between the first received bid and the second received bid is less than the predetermined bid spread, withhold running the sample through the set of cascading classifier models.

20. The non-transitory computer readable medium of claim 16 , wherein each subsequent classifier model in the set of cascading classifier models is more restricting than a prior.