System and method using sampling for scheduling advertisements in an online auction with budget and time constraints

1. A computer system for an online advertising auction, comprising:

an input/output interface receiving, from each of a plurality of advertisers bidding for advertising slots in the online advertising auction:

a bid for an advertising slot based on a keyword, wherein said bid comprises:

an arrival time at the online advertising auction;

a departure time at the online advertising auction;

a budget to be spent between the arrival time and the departure time, wherein said budget may not be exceeded; and

a value per click;

wherein it is assumed that the budget is greater than the value per click;

a processor device configured as a multi-armed bandit engine learning a valuation of advertisements through sampling by scheduling the advertisements for web page placements in an online advertising auction with budget and time constraints to optimize payments for maximizing welfare of advertisers by performing:

computing critical values for each of the plurality of advertisers in a given time period of the online advertising auction, wherein a critical value is a minimum value for the advertising slot in the given time period;

computing an interim price based upon the critical values for each of the plurality of advertisers in the online advertising auction;

charging the plurality of advertisers departing from the online advertising auction the computed interim price; and

updating a remaining budget of each of the plurality of advertisers in the online advertising auction; and

a storage device operably coupled to the multi-armed bandit engine storing a plurality of budgets and a plurality of bids each associated with an advertisement allocated to web page placements in the online advertising auction with budget and time constraints;

wherein the computer system selects a plurality of advertisers for sampling advertisements in the online advertising auction and samples advertisements for a subset of the plurality of advertisers.

2. The system of claim 1 further comprising a model generator for creating a multi-armed bandit model used by the multi-armed bandit engine.

3. The system of claim 1 further comprising a payoff optimizer operably coupled to the multi-armed bandit engine for optimizing payments for the advertisements sampled in the online advertising auction with budget and time constraints to maximize the welfare of the advertisers.

4. The system of claim 1 wherein computing the critical values comprises:

selecting time periods occurring since the arrival times of the plurality of advertisers in the online advertising auction;

initializing status set parameters to zero for the advertisers entering and departing the online auction during the selected time periods;

selecting a subset of the advertisers for sampling from among the advertisers entering and departing the online auction during the selected time periods;

collecting status set parameters from the subset of advertisers; and

setting the critical values for the subset of advertisers based on the collected status set parameters.

5. The system of claim 4 wherein collecting the status set parameters comprises:

allocating an advertising slot for each advertiser;

monitoring the advertising slot for each of the selected time periods;

determining a click-through status as a sum of clicks received during each of the selected time periods;

determining the critical value as a minimum sum of values that could have resulted in the click-through status, wherein the values comprise:

a first value representing a value for the advertising slot when the advertiser did not gain an impression during a time period;

a second value representing a value for the advertising slot when the advertiser had an impression but not a click through during a time period; and

a third value representing a value for the advertising slot when the advertiser had an impression and a click-through during a time period.

6. The system of claim 4 wherein selecting the subset of advertisers comprises:

determining the advertisers currently in the online advertising auction; and

removing the advertisers with a payoff rate lower than a threshold.

7. The system of claim 4 wherein initializing the status set parameters comprises:

initializing the click-through rate for an advertiser to zero;

initializing the click-through status for an advertiser to zero; and

initializing the payoff rate per click to zero.

8. The system of claim 5 wherein determining the critical value further comprises:

determining a click-through rate as a probability of receiving a click when the advertiser is allotted an impression.

9. The system of claim 8 further comprising estimating an accuracy of the click-through rate based on a number of impressions received by the advertiser.

10. The system of claim 5 wherein determining the critical value further comprises:

incrementing a payoff rate when receiving a click; and

decrementing the payoff rate when not receiving a click.

11. The system of claim 5 wherein the sum of values represents a union of the first and third values.