Competition control method, system and storage medium

Abstract

The present disclosure proposes a competition control method, system and storage medium, which relate to the field of computer technology. A competition control method disclosed in the present disclosure includes: obtaining the real-time estimated conversion parameters of advertisements through an estimation model; determining the traffic quality layer to which the real-time consumption at the corresponding moment of the real-time estimated conversion parameters belongs according to the real-time estimated conversion parameters and the estimated conversion parameter stratification threshold; for each traffic quality layer, adjusting the competition probability of each traffic quality layer according to the difference between the actual consumption in the previous time window and the target consumption in the subsequent time window in two adjacent time windows; and conducting advertisement competition according to the adjusted competition probability. Through such a method, the generation of burst traffic can be avoided and the stability of the competing traffic can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a competition control method, system and storage medium. Background Art

[0002] The online advertising industry is a booming new industry in recent years, with a vast market of billions of dollars worldwide. Since the development of online advertising systems, RTB (Real Time Bidding) advertising has occupied a large part of the market share, and the technology in this field is also constantly developing and innovating. However, due to the large number of advertisers and the diverse needs, how to do RTB advertising well is still a major problem facing the industry.

[0003] RTB advertisers pursue advertising effectiveness and experience based on a certain budget. Summary of the invention

[0004] One purpose of the present disclosure is to improve the stability of competing traffic.

[0005] According to one aspect of some embodiments of the present disclosure, a competition control method is proposed, including: obtaining real-time estimated conversion parameters of advertisements through an estimation model; determining, based on the real-time estimated conversion parameters and the estimated conversion parameter stratification thresholds, the traffic quality layer to which the real-time consumption at the moment corresponding to the real-time estimated conversion parameters belongs; for each traffic quality layer, adjusting the competition probability of each traffic quality layer based on the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows; and conducting advertisement competition based on the adjusted competition probability.

[0006] In some embodiments, adjusting the probability of each traffic quality layer participating in the competition includes: obtaining the consumption error between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows; when the consumption error belongs to the predetermined first interval, the probability of participating in the competition is not adjusted; when the consumption error does not belong to the predetermined first interval, the probability of participating in the competition of each traffic quality layer is adjusted in sequence according to the predetermined adjustment strategy in the order of the estimated conversion interval thresholds corresponding to the traffic quality layer.

[0007] In some embodiments, obtaining the consumption error between the actual consumption of an earlier time window and the target consumption of a later time window in two adjacent time windows includes: adjusting the actual consumption of the earlier time window according to the ratio of the time length of the later time window to the earlier time window to obtain the proportional change consumption; obtaining the difference between the target consumption of the later time window and the proportional change consumption as the consumption error.

[0008] In some embodiments, when the consumption error does not belong to the predetermined first interval, the probability of participating in the competition of each traffic quality layer is adjusted in sequence according to the size order of the estimated conversion interval thresholds corresponding to the traffic quality layers in accordance with the predetermined adjustment strategy, including: when the consumption error is less than the low threshold of the predetermined first interval, the probability of participating in the competition of each traffic quality layer is reduced in sequence from low to high according to the estimated conversion interval thresholds corresponding to the traffic quality layers; and when the consumption error is greater than the high threshold of the predetermined first interval, the probability of participating in the competition of each traffic quality layer is increased in sequence according to the estimated conversion interval thresholds corresponding to the traffic quality layers from high to low.

[0009] In some embodiments, adjusting the probability of participating in competition of each traffic quality layer according to a predetermined adjustment strategy includes: for each traffic quality layer, obtaining the sum of the consumption error and the actual consumption of the previous time window of the corresponding traffic quality layer, and determining the adjusted consumption; adjusting the probability of participating in the previous time window according to the ratio of the adjusted consumption to the actual consumption of the previous time window of the corresponding traffic quality layer, and obtaining the probability of participating in the subsequent time window.

[0010] In some embodiments, the competition control method also includes: when the actual consumption of a traffic quality layer is less than a predetermined first consumption threshold, increasing the advertising bid by a predetermined percentage, and updating the competition probability of each traffic quality layer until the actual consumption of each traffic quality layer is not less than the predetermined first consumption threshold.

[0011] In some embodiments, the competition control method also includes: before determining to adjust the competition probability of each traffic quality layer, determining whether data flow congestion occurs according to the change in the cumulative value of actual consumption; when the difference between the latest cumulative value of actual consumption and the cumulative value of the previous predetermined number of actual consumption is less than or equal to the congestion threshold, determining that data flow congestion occurs; otherwise, performing the operation of adjusting the competition probability of each traffic quality layer for each traffic quality layer according to the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows.

[0012] In some embodiments, the competition control method also includes: determining the target consumption in the subsequent time window based on the expected consumption value, the actual cumulative consumption ratio in the previous time window and the expected cumulative consumption ratio in the subsequent time window, or based on the distribution of expected cumulative consumption, expected unit advertising cost and actual unit advertising cost.

[0013] In some embodiments, the target consumption in the subsequent time window is determined based on the expected consumption value, the percentage of actual cumulative consumption in the previous time window and the percentage of expected cumulative consumption in the subsequent time window, or based on the distribution of expected cumulative consumption, expected single advertisement cost and actual single advertisement cost, including: when the actual conversion value is greater than the predetermined first conversion value, if the current actual single advertisement cost on that day is not lower than the expected single advertisement cost, then the target consumption in the subsequent time window is determined based on the distribution of expected cumulative consumption, expected single advertisement cost and actual single advertisement cost.

[0014] In some embodiments, according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost, determining the target consumption of the subsequent time window further includes:

[0015] In the case where the actual conversion value is greater than the predetermined first conversion value, if the actual single advertisement cost on that day is lower than the expected single advertisement cost, then

[0016] Determine an expected consumption value according to an expected single advertisement cost and an actual conversion value, and determine a first consumption in a subsequent time window according to the expected consumption value, a proportion of actual cumulative consumption in a previous time window, and a proportion of expected cumulative consumption in a subsequent time window;

[0017] Determine the second consumption in the later time window according to the expected consumption distribution, the expected single advertisement cost and the actual single advertisement cost;

[0018] The larger one of the first consumption and the second consumption is determined as the target consumption in the subsequent time window.

[0019] In some embodiments, determining the target consumption in the subsequent time window based on the expected consumption value, the percentage of actual cumulative consumption in the prior time window and the percentage of expected cumulative consumption in the subsequent time window, or based on the distribution of expected cumulative consumption, expected single advertisement cost and actual single advertisement cost also includes: when the actual conversion value is not greater than the predetermined first conversion value, if there is actual consumption data for a historical date, taking the weighted sum of the actual cumulative consumption in the prior time window of the day and the actual cumulative consumption at the corresponding moment of the previous day as the expected consumption value; determining the target consumption in the subsequent time window based on the expected consumption value, the percentage of actual cumulative consumption on the day and the percentage of expected cumulative consumption in the subsequent time window.

[0020] In some embodiments, the competition control method also includes: when the actual conversion value is not greater than the predetermined first conversion value, if there is no actual consumption data for the historical date, the product of the current actual cumulative consumption ratio on the day and the expected full-day consumption based on the expected single advertising cost is used as the target consumption.

[0021] In some embodiments, the competition control method also includes: determining the expected cumulative consumption in the subsequent time window based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio of the current date, and the total target consumption of the current date.

[0022] In some embodiments, the competition control method further includes: determining a target total consumption value for the current date based on the actual total consumption value of historical dates, the expected single advertisement cost, and the actual single advertisement cost.

[0023] In some embodiments, determining the expected cumulative consumption in the subsequent time window includes: determining the value to be consumed on the current date based on the actual cumulative consumption ratio of the current date and the target total consumption of the current date; distributing the value to be consumed on the current date in the remaining period of the current date according to the distribution of actual consumption on historical dates, and determining the expected cumulative consumption in the subsequent time window.

[0024] In some embodiments, determining the target total consumption value includes: when the actual conversion value of a historical date is greater than or equal to a predetermined second conversion value, adjusting the actual total consumption value of the corresponding historical date according to the ratio of the difference between the expected single advertisement cost and the actual single advertisement cost to the expected single advertisement cost, and obtaining a first total consumption value; when the actual conversion value of the corresponding historical date is less than the predetermined second conversion value, determining the actual total consumption value of the corresponding historical date as the first total consumption value; multiplying the first total consumption value and the expected single advertisement cost with the corresponding hyperparameters, respectively, and comparing them with the predetermined minimum target to determine the largest item as the second total consumption value of the current date; when the number of historical dates is greater than 1, obtaining the weighted sum of the second total consumption values ​​of different historical dates as the target total consumption value of the current date.

[0025] In some embodiments, determining the target consumption total value based on the actual consumption total value of historical dates, the expected single advertisement cost and the actual single advertisement cost also includes: when the number of historical dates is equal to 1, using the second consumption total value as the target consumption total value for the current date.

[0026] In some embodiments, the competition control method also includes: dividing the estimated conversion parameters into a predetermined number of equal parts according to the distribution of the estimated conversion parameters of historical dates of a predetermined number of days and a predetermined number of layers, and obtaining the quantile parameters for dividing the traffic quality layers as the estimated conversion parameter stratification threshold.

[0027] In some embodiments, the competition control method also includes at least one of the following: at the beginning of each day, the competition probabilities of all traffic quality layers of all advertising units are initialized to a predetermined first initial value; in the case of a newly added advertising unit, the competition probabilities of all traffic quality layers of the newly added advertising unit are initialized to a predetermined second initial value; when the target consumption in the subsequent time window is less than or equal to a predetermined second consumption threshold, the competition probability of the traffic quality layer with the highest threshold of the estimated conversion interval is set to the predetermined first probability, and the competition probabilities of other traffic quality layers are set to be less than the predetermined first probability; or after each adjustment of the competition probability of each traffic quality layer, the competition probability of the layer with the highest estimated conversion interval threshold corresponding to each traffic quality layer with a competition probability of 0 is adjusted to a predetermined second probability.

[0028] In some embodiments, the competition control method further includes: when the actual consumption in the previous time window is less than or equal to the predetermined second consumption threshold, if the consumption error does not belong to the predetermined first interval, modifying the actual consumption in the previous time window to the predetermined third consumption.

[0029] In some embodiments, the upper limit of the competition probability is 1, and the lower limit is 0.

[0030] According to one aspect of some embodiments of the present disclosure, a competition control system is proposed, including: an estimation unit, configured to obtain real-time estimated conversion parameters of advertisements through an estimation model; a stratification unit, configured to determine, based on the real-time estimated conversion parameters and the estimated conversion parameter stratification threshold, the traffic quality layer to which the real-time consumption at the moment corresponding to the real-time estimated conversion parameters belongs; a probability adjustment unit, configured to adjust, for each traffic quality layer, the competition probability of each traffic quality layer based on the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows; and a competition unit, configured to conduct advertisement competition based on the adjusted competition probability.

[0031] In some embodiments, the competition control system also includes a blocking determination unit, which is configured to: before determining to adjust the competition probability of each traffic quality layer, determine whether data flow blocking occurs based on the change in the cumulative value of actual consumption; when the difference between the latest cumulative value of actual consumption and the cumulative value of a previous predetermined number of actual consumption is less than or equal to a blocking threshold, determine that data flow blocking occurs; otherwise, activate the probability adjustment unit.

[0032] In some embodiments, the competition control system also includes: a window target consumption determination unit, which is configured to determine the target consumption of the subsequent time window based on the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or based on the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost.

[0033] In some embodiments, the competition control system also includes: an expected cumulative consumption determination unit, which is configured to determine the expected cumulative consumption in the subsequent time window based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio of the current date and the target consumption total value of the current date.

[0034] In some embodiments, the competition control system further includes: a target consumption total value determination unit configured to determine the target consumption total value of the current date based on the actual consumption total value of the historical date, the expected unit advertisement cost and the actual unit advertisement cost.

[0035] In some embodiments, the competition control system also includes: a stratification threshold determination unit, which is configured to divide the estimated conversion parameters into a predetermined number of equal parts according to the distribution of the estimated conversion parameters on historical dates of a predetermined number of days and a predetermined number of layers, and obtain the quantile parameters of the divided traffic quality layers as the estimated conversion parameter stratification threshold.

[0036] In some embodiments, the competition control system also includes a first processing unit, and is further configured to perform at least one of the following: at the beginning of each day, the competition probabilities of all traffic quality layers of all advertising units are initialized to a predetermined first initial value; in the case where a new advertising unit is added, the competition probabilities of all traffic quality layers of the new advertising unit are initialized to a predetermined second initial value; when the target consumption in the subsequent time window is less than or equal to a predetermined second consumption threshold, the competition probability of the traffic quality layer with the highest threshold of the estimated conversion interval is set to a predetermined first probability, and the competition probabilities of other traffic quality layers are set to be less than the predetermined first probability; or after each adjustment of the competition probability of each traffic quality layer, the competition probability of the layer with the highest estimated conversion interval threshold corresponding to each traffic quality layer with a competition probability of 0 is adjusted to a predetermined second probability.

[0037] According to one aspect of some embodiments of the present disclosure, a competition control system is proposed, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the above competition control methods based on instructions stored in the memory.

[0038] According to one aspect of some embodiments of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the steps of any one of the above-mentioned competing control methods are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0040] Figure 1The flowcharts are of some embodiments of the competition control method disclosed in the present invention.

[0041] Figure 2 The present invention is a flowchart of some embodiments of adjusting the contention probability in the contention control method of the present invention.

[0042] Figure 3 Flowcharts of other embodiments of the competition control method disclosed in the present invention.

[0043] Figure 4 Flow chart of some embodiments of target consumption determination in the competing control method of the present disclosure.

[0044] Figure 5 Flowcharts of some embodiments of determining expected cumulative consumption in the competing control method of the present disclosure.

[0045] Figure 6 Flowchart of some embodiments of determining the target total consumption value in the competing control method disclosed in the present invention.

[0046] Figure 7 Schematic diagrams of some embodiments of the competition control system disclosed in the present invention.

[0047] Figure 8 Schematic diagrams of other embodiments of the competition control system disclosed in the present invention.

[0048] Fig. 9 Schematic diagrams of some further embodiments of the competition control system disclosed herein. DETAILED DESCRIPTION

[0049] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments.

[0050] In related technologies, the advertising effect is often ensured by accelerating the broadcast at night, or by dividing the budget into time periods to allocate the budget consumption within a period of time, or by using the PID (Proportion Integral Differential) feedback adjustment algorithm originally introduced from the field of ship autopilots.

[0051] The technology used by the industry to accelerate playback at night to ensure that the budget for the day is consumed may result in the concentration of bids at night to obtain some low-quality traffic at high prices at night, resulting in the advertising effect failing to meet advertisers' expectations, and may also lead to a surge in network pressure; in addition, the probability of traffic blockage is greatly increased, which puts great pressure on both the advertising platform and the advertiser's platform, and at the same time, for users, it will cause a decrease in access speed, reducing user experience. The method of dividing the budget set by the advertiser into time periods can guarantee a certain degree of uniform consumption of the budget throughout the day, but it requires a lot of manual operation and experience accumulation, and there is a certain time delay in taking effect, and there is a risk of not tracking changes in the traffic environment. Although the PID feedback adjustment algorithm can adjust the advertising effect in a relatively timely manner, it cannot avoid extreme situations such as short-term bursts of advertising.

[0052] Related technologies propose a budget control strategy based on the probability of participation (pacing rate), introduce KPI (Key Performance Indicator) for optimization, and achieve statistical completion of the interval estimation of the KPI quantiles of the current period. Within the interval, the KPI is bid according to the pacing_rate. If it is higher than the upper limit, it must bid, and if it is lower than the lower limit, it will not bid. However, there are many parameters that need to be estimated, and the pacing rate has the disadvantage of being too unstable due to large fluctuations. The fluctuation of pacing_rate will directly affect the fluctuation of traffic, resulting in unstable traffic and unstable network.

[0053] The flowcharts of some embodiments of the competition control method disclosed in the present invention are as follows: Figure 1 shown.

[0054] In step 130, the real-time estimated conversion parameters of the advertisement are obtained through the estimation model. In some embodiments, an estimation model can be constructed to predict the advertisement effect of the current request in real time as the real-time estimated conversion parameters. The real-time estimated conversion parameters can be indicators that advertisers care about, such as PCVR (Predict Conversion Rate) or PCTR (Predict ClickThrough Rate).

[0055] In some embodiments, the estimated model structure may vary with actual conditions such as the amount of training data, actual business accuracy requirements, and time requirements.

[0056] In some embodiments, each time actual consumption occurs (ie, an advertisement is bid for), an estimation operation is triggered.

[0057] In step 140, the traffic quality layer to which the real-time consumption at the time corresponding to the real-time estimated conversion parameter belongs is determined according to the real-time estimated conversion parameter and the estimated conversion parameter stratification threshold.

[0058] In some embodiments, the real-time estimated conversion parameters may be determined after actual consumption occurs, and the traffic quality layer may be determined based on the real-time estimated conversion parameters. Then, the actual consumption belongs to the traffic quality layer determined above.

[0059] In step 160, for each traffic quality layer, the probability of participating in competition of each traffic quality layer is adjusted according to the difference between the actual consumption in the previous time window and the target consumption in the subsequent time window in two adjacent time windows.

[0060] In step 170, advertisement bidding is performed according to the adjusted bidding probability.

[0061] Through this method, the actual consumption can be processed in layers according to the real-time estimated conversion parameters, and the probability of each traffic quality layer participating in the competition can be adjusted with the time window as the time processing unit according to the consumption of different layers and the target consumption, thereby improving the differentiated processing capabilities for different quality traffic. Since different traffic qualities have different degrees of influence on subsequent traffic fluctuations, the impact of high traffic quality consumption on traffic fluctuations is more obvious, so that differentiated processing can be carried out in time to avoid the generation of sudden traffic, improve the stability of competing traffic, and at the same time help to ensure user access speed and user experience.

[0062] In some embodiments, Figure 1 As shown, between step 140 and step 160, step 150 may be further included: determining whether data flow blocking occurs. If data flow blocking does not occur, step 160 is executed; if data flow blocking occurs, step 151 is executed.

[0063] In some embodiments, whether data flow congestion occurs can be determined based on the change in the accumulated value of actual consumption; when the change in the latest accumulated value of actual consumption and the second latest accumulated value of actual consumption is less than or equal to a congestion threshold, it is determined that data flow congestion occurs; otherwise, it is determined that no congestion currently occurs.

[0064] In some embodiments, whether a blockage occurs may be determined by:

[0065] a. Maintain a one-dimensional array, [(s1,tm1), (s2,tm2), ..., (s n ,tm n )], where n is a hyperparameter, indicating the maximum number of elements in the recorded array. n is the sum of consumption from the 1st to the nth in the time period, tm nThe timestamp of the nth consumption;

[0066] b. Calculate the latest consumption and S of all ad units in the time period obtained this time, and record the current timestamp T M Ad unit refers to the smallest unit of advertising delivery, which can be used to count the number of exposure clicks and the number of orders brought in to measure the effectiveness of the delivery.

[0067] c. Update the array, [(s2,tm2),…,(s n ,tm n ), (S, TM)] and sort them according to timestamp tm;

[0068] d. Count from the last bit of the array and determine whether it exists (Ss i )>interv;

[0069] Where i = (n, n-1, ..., 1), interv is a hyperparameter, indicating the blocking threshold. If this inequality holds, it is determined that the data is not blocked, and the subsequent steps of updating the probability of participating in the competition are performed normally; if this inequality still does not hold after traversing the array, it is determined that the data flow is blocked.

[0070] In step 151, a flow congestion alarm is issued so that emergency measures can be automatically or manually executed. In some embodiments, the probability of participating in the competition can be kept unchanged, and the system can directly exit and send an alarm.

[0071] Through this method, data flow blockage can be discovered and handled in a timely manner, thereby improving the operational reliability of advertising competition, avoiding incorrect adjustments to the competition probability caused by data flow blockage, and improving the reliability of competition control.

[0072] In some embodiments, to facilitate data management, in each advertising unit hierarchical data structure statistics module, for each advertising unit, a hierarchical data structure can be maintained, and each layer corresponds to an advertising request group. Each advertising request group can include a real-time budget allocation value, a conversion rate PCVR distribution interval, consumption in the latest time period, and data blocking judgment, wherein the real-time budget allocation value is the calculation result produced by the hourly consumption control, and the initial value of the time-divided budget at the advertising unit granularity is obtained.

[0073] In the time window actual consumption statistics step, the actual consumption of the previous time window, the time span of the previous time window, the target consumption of the next time window, and the time span of the next time window are counted. Taking 0:00 as the initialization time of each day as an example, the actual consumption of the previous window refers to the consumption from 0:00 to the current time when it is initialized at 0:00 each day, and refers to the consumption from the timestamp after the last update to the current time when it is running normally. The time span of the previous window refers to the time span between the timestamp of the latest data that can be obtained and the initialization time (0:00 that has passed) when it is initialized at 0:00 each day, and refers to the time span between the timestamp of the latest data that can be obtained and the timestamp of the end of the last program run when it is running normally. The target consumption of the next window refers to the difference between the cumulative target consumption to the end time of the next window and the current actual cumulative consumption. The time span of the next window is the length of time from the current time to the end point of the next nearest window (if in hours, it is the next full hour).

[0074] In some embodiments, since the estimated value of the model is affected by many factors and cannot be guaranteed to be 100% accurate, in the above step 130, after obtaining the real-time preliminary estimated conversion parameters directly based on the estimation model, the current real-time preliminary estimated conversion parameters can be corrected based on the difference between the historical actual conversion parameters and the estimated conversion parameters of the corresponding historical time period. Taking PCVR as an example, the difference between the actual CVR (Convertion Rate, conversion probability) based on history and the estimated PCVR of the corresponding historical time period is used to correct the current PCVR. The statistical frequency can be set according to the business situation (such as hourly level), and the formula can be as follows:

[0075] pcvr_use = pcvr_predict * (cvr_hour / pcvr_hour)

[0076] Among them, pcvr_predict is the output value estimated by the model, that is, the real-time preliminary estimated conversion parameter, cvr_hour is the credible cvr value within a period of history, pcvr_hour is the model prediction mean value within the corresponding period of history, and pcvr_use is the corrected current request estimate value, that is, the real-time estimated conversion parameter determined in step 130, which is used to perform subsequent processing.

[0077] Through such a method, the deviation of the estimation model can be corrected and the accuracy of the actual consumption stratification can be improved, thereby ensuring that the advertisers' requirements for conversion effects can be met while improving the smoothness of traffic, which is conducive to the promotion and application of the competitive control scheme of this application for improving the smoothness of traffic.

[0078] In some embodiments, when calculating cvr_hour, it can be calculated at the ad unit granularity and a click threshold can be set. If the amount of data at the ad unit granularity is insufficient, it can be sequentially fallen back to, for example, by advertiser + third-level category. Third level category The order of overall granularity is backed off to ensure that the amount of click data reaches a credible and convincing threshold, thus ensuring the credibility of the data.

[0079] In some embodiments, when calculating pcvr_hour, it can be calculated at the ad unit granularity and a click threshold can be set. If the ad unit granularity is insufficient to use the data of the last hour, it will be prioritized to fall back to the 2-hour data until it falls back to the last moment before the model was updated yesterday (to ensure that the estimated value of the same model is corrected to avoid errors caused by model updates), and then use the data by advertiser + third-level category. Third level category Logical fallback for the order of overall granularity.

[0080] Through this method, it can be ensured that the amount of data being processed is of universal significance, avoiding the impact of occasional data and improving the reliability of processing.

[0081] In some embodiments, the estimated conversion parameter stratification threshold can be determined based on the analysis of historical data. In some embodiments, the competition control method further includes: dividing the estimated conversion parameter into a predetermined number of equal parts according to the distribution of the estimated conversion parameters on historical dates of a predetermined number of days and a predetermined number of strata, and obtaining the quantile parameters for dividing the traffic quality layers as the estimated conversion parameter stratification threshold. Through such a method, it can be ensured that the stratification of the traffic quality layers can match the actual conversion situation of the advertising unit, thereby improving the accuracy and response efficiency of traffic smoothness control.

[0082] In some embodiments, flowcharts of some embodiments of the competition control method corresponding to the above step 160 are as follows: Figure 2 shown.

[0083] In step 261, the consumption error between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows is obtained.

[0084] In some embodiments, the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows may be directly calculated as the consumption error.

[0085] In some embodiments, considering that the lengths of the preceding and succeeding time windows may be inconsistent, the window lengths are normalized, and the consumption error is calculated according to the following formula:

[0086] residual = c next –c last *(tm next / tm last )

[0087] Among them, residual is the consumption error, c next is the target consumption in the later time window, c last is the actual consumption in the previous time window, tm next is the length of the subsequent time window, tm last is the length of the previous time window.

[0088] In step 262, it is determined that the consumption error belongs to a predetermined first interval. In some embodiments, the predetermined first interval may be a numerical interval close to the numerical value 0, for example, the predetermined first interval is 0 itself, or [-0.01, 0.01], etc. The predetermined first interval may be adjusted as needed.

[0089] If the consumption error belongs to the predetermined first interval, it means that the actual consumption is close to the expected one, and step 263 is executed; otherwise, step 264 is executed.

[0090] In step 263, the competition probability is not adjusted.

[0091] In step 264, it is determined whether the consumption error is less than a lower threshold of a predetermined first interval. If the consumption error is less than the lower threshold of the predetermined first interval, it means that the actual consumption is faster than expected, and step 265 is executed; otherwise, it means that the actual consumption is slower than expected, and step 266 is executed.

[0092] In step 265, the probability of participating in each traffic quality layer is reduced in order from low to high according to the estimated conversion interval thresholds corresponding to the traffic quality layers. In some embodiments, the lower limit of the adjustment of the probability of participating in the competition is 0. If the probability of participating in the traffic quality layer being adjusted is already 0, the layer is skipped.

[0093] In step 266, the competition probability of each traffic quality layer is increased in order from high to low according to the estimated conversion interval threshold corresponding to the traffic quality layer. In some embodiments, the upper limit of the adjustment of the competition probability is 1. If the competition probability of the traffic quality layer being adjusted is already 1, then the layer is skipped.

[0094] In some embodiments, for determining the new contention probability in the above steps 265 and 266, the sum of the consumption error and the actual consumption of the previous time window of the corresponding traffic quality layer may be obtained for each traffic quality layer to determine the adjusted consumption. Further, according to the ratio of the adjusted consumption to the actual consumption of the previous time window of the corresponding traffic quality layer, the contention probability of the previous time window is adjusted to obtain the contention probability of the subsequent time window.

[0095] In some embodiments, the adjustment may be made based on the formula:

[0096] p next (l)=p last (l) * (c last (l)+residual) / c last (l)

[0097] In the formula, p next (l) is the probability of participating in the competition of the lth traffic quality layer in the later time window, p last (l) is the probability of participating in the competition of the first layer in the previous time window, residual is the consumption error, c last (l) is the actual consumption of the lth layer in the previous time window, l is the flow quality layer identifier, a positive integer, and its value range is within the range of the number of flow quality layers.

[0098] Through such a method, real-time adjustments can be made according to the difference between consumption and expected consumption, avoiding traffic surges caused by too fast consumption and subsequent traffic increases caused by too slow consumption, thereby improving traffic smoothness; giving priority to reducing the probability of low-quality traffic quality layers participating in the competition, and giving priority to increasing the probability of high-quality traffic quality layers participating in the competition, which can reduce the impact of adjustments on advertisers' businesses and improve the efficiency of adjustments on traffic smoothness.

[0099] In some embodiments, the number of traffic quality layers may be initialized to L=n, and each layer shares a global contention probability as a hyperparameter G. Each layer in the first time window will generate a certain consumption, which can serve as the basis for subsequent contention probability adjustment.

[0100] When the actual consumption of a traffic quality layer is less than the predetermined first consumption threshold, the ad bid is increased by a predetermined percentage, and the bid probability of each traffic quality layer is updated until the actual consumption of each traffic quality layer is not less than the predetermined first consumption threshold. For example, if the bid of an ad unit is low, resulting in no consumption in some layers, the bid of the ad unit can be increased by a certain percentage until all layers have consumption and no longer increase. Each adjustment will save the latest bid probability of each layer with non-zero consumption, which can be used later. Figure 2 The basis for adjusting the probability of participating in the competition is as shown.

[0101] This method can avoid the situation where the initial probability of competition is 0, resulting in no actual consumption in the future and falling into an infinite loop, thereby improving the robustness of competition control.

[0102] In some embodiments, corresponding processing may be performed for some special cases.

[0103] In some embodiments, the competition probabilities of all traffic quality layers of all advertising units are initialized to a predetermined first initial value at the beginning of each day. For example, when initialized at 0:00 each day, the competition probabilities of all advertising units are initialized to the hyperparameter G1, thereby avoiding data deviation caused by daily processing and ensuring operational stability.

[0104] In some embodiments, when there are newly added advertising units, the probability of participating in competition of all traffic quality layers of the newly added advertising units is initialized to a predetermined second initial value. For example, the probability of participating in competition of each layer of the newly added advertising units on that day is initialized to the hyperparameter G2, and the subsequent first adjustment is made based on G2, thereby enabling automatic operation when new advertising units are added.

[0105] In some embodiments, when the target consumption in the subsequent time window is less than or equal to the predetermined second consumption threshold, the competition probability of the traffic quality layer with the highest threshold value in the estimated conversion interval is set to the predetermined first probability, and the competition probability of other traffic quality layers is set to be less than the predetermined first probability. For example, if the target consumption of the next window is less than or equal to 0, the competition probability is set to 0.01 for the highest layer and 0 for other layers. Through such a method, while reasonably avoiding overspending, it is possible to avoid the system from falling into a state where the competition probability of each layer is 0, thereby improving the robustness of the system.

[0106] In some embodiments, after each adjustment of the competition probability of each traffic quality layer, the competition probability of the layer with the highest estimated conversion interval threshold corresponding to each traffic quality layer with a competition probability of 0 is adjusted to a predetermined second probability, for example, the highest PCVR layer with a pacing rate of 0 is set to 0.1 for accumulating data.

[0107] In some embodiments, when the actual consumption of the previous time window is less than or equal to the predetermined second consumption threshold, if the consumption error does not belong to the predetermined first interval, the actual consumption of the previous time window is modified to the predetermined third consumption. For example, if the previous window consumption of a certain layer is 0, and the competition probability of the layer needs to be adjusted (the consumption error is not within the predetermined first interval), the previous window consumption is set to the default hyperparameter F. Such a method can avoid the situation where the subsequent competition probability changes too much when the actual consumption of the previous window is too small, and further improve the stability of the traffic.

[0108] In some embodiments, the target consumption in the later time window may be a specified value based on experience or statistics, thereby reducing system pressure and increasing processing speed.

[0109] In other embodiments, the target consumption in the later time window is calculated based on the expected consumption value, the percentage of actual cumulative consumption in the previous time window and the percentage of expected cumulative consumption in the later time window, or based on the distribution of expected cumulative consumption, expected unit advertisement cost and actual unit advertisement cost.

[0110] Through such a method, the target consumption in the subsequent time window can be dynamically determined based on historical data and real-time actual consumption, thereby improving the reliability of the determined target consumption in the subsequent time window and further improving the robustness of the flow smoothness control effect.

[0111] In some embodiments, the expected cumulative consumption to be used in determining the target consumption in the later time window may be a specified value based on experience or statistics, thereby reducing system pressure and increasing processing speed.

[0112] In other embodiments, the expected cumulative consumption in the subsequent time window can be determined based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio on the current date, and the total target consumption on the current date. In this way, the expected cumulative consumption can be dynamically determined based on historical data and actual consumption, thereby improving the reliability of the expected cumulative consumption and further improving the robustness of the effect of flow smoothness control.

[0113] In some embodiments, the target total consumption value for the current date required in determining the expected cumulative consumption process may be a specified value based on experience or statistics, thereby reducing system pressure and increasing processing speed.

[0114] In other embodiments, the target total consumption value of the current date can be determined based on the actual total consumption value of the historical date, the expected single advertisement cost, and the actual single advertisement cost. By this method, the target total consumption value of the current date can be dynamically determined, the reliability of the target total consumption value of the current date is improved, and the robustness of the effect of flow smoothness control is further improved.

[0115] Flow charts of other embodiments of the competition control method disclosed in the present invention are as follows: Figure 3 shown.

[0116] In step 311, the target total consumption value for the current date is determined based on the actual total consumption value of the historical date, the expected single advertisement cost, and the actual single advertisement cost.

[0117] In step 312, the expected cumulative consumption in the subsequent time window is determined based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio on the current date, and the target total consumption on the current date.

[0118] In step 320, the target consumption in the subsequent time window is determined based on the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or based on the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost.

[0119] Through this method, the target total consumption of the current date, the expected cumulative consumption and the target consumption in the subsequent time window can be gradually and dynamically determined, so that the adjustment of the competition probability conforms to the real-time situation, further improving the robustness of the effect of traffic smoothness control.

[0120] In some embodiments, the flowcharts of some embodiments of the above step 320 are as follows: Figure 4 shown.

[0121] In step 4201, the actual conversion value of the day is obtained. The actual conversion value of the day refers to the actual conversion value of the current date from the start time to the current time.

[0122] In step 4202, it is determined that the actual conversion value is greater than the predetermined first conversion value. In some embodiments, the predetermined first conversion value may be set as a confidence threshold value. If it is greater than the confidence threshold value, the current actual conversion value data is credible data; otherwise, the data is considered to be too small and is an unconfident value. In some embodiments, the predetermined first conversion value may be

[0123] If the actual conversion value is greater than the predetermined first conversion value, execute step 4203; otherwise, execute step 4208.

[0124] In step 4203, it is determined whether the actual cost per advertisement on the current day is lower than the expected cost per advertisement. If it is lower than the expected cost per advertisement, step 4204 is executed; otherwise, step 4207 is executed.

[0125] In some embodiments, the actual cost of a single advertisement at the current time of the day can be calculated by the ratio of the actual cumulative consumption at the current time of the day to the sum of the total number of clicks on each advertisement.

[0126] In step 4204, the expected consumption value is determined based on the expected single advertisement cost and the actual conversion value, and the first consumption in the subsequent time window is determined based on the expected consumption value, the actual cumulative consumption ratio of the previous time window, and the expected cumulative consumption ratio of the subsequent time window.

[0127] In some embodiments, the expected consumption value reasonable_cost = tcpa * order, where tcpa is the expected single advertisement cost, that is, the order line cost expected by the advertiser, and order is the actual conversion value. In some embodiments, since the price of an advertisement unit can be calculated based on the number of clicks, that is, a unit price is charged per click, the expected single advertisement cost is the expected cost of the advertisement being clicked once.

[0128] In some embodiments, the first consumption in the subsequent time window may be determined according to the following formula:

[0129] target_cost_1 = reasonable_cost / ratio_1* ratio_2,

[0130] Among them, target_cost_1 is the first consumption in the later time window, ratio_1 is the actual cumulative consumption ratio in the first time window, and ratio_2 is the expected cumulative consumption ratio in the later time window.

[0131] In step 4205, the second consumption in the subsequent time window is determined based on the expected consumption distribution, the expected single advertisement cost, and the actual single advertisement cost.

[0132] In some embodiments, the second consumption in the subsequent time window may be determined according to the following formula:

[0133] target_cost_2 = old_target_cost * (1 - UPDATE_RATE * cpa_diff),

[0134] cpa_diff = (tcpa - cpa) / cpa

[0135] Among them, target_cost_2 is the second consumption in the later time window; old_target_cost is the cumulative target consumption ratio of the expected cutoff point of the previous time window, and the cumulative target consumption ratio refers to the cumulative value of the target consumption ratio from the start time of the corresponding date to the target time. cpa_diff is the difference between the expected single advertisement cost and the actual single advertisement cost, tcpa is the expected single advertisement cost, cpa is the actual single advertisement cost, and UPDATE_RATE is a hyperparameter, indicating the update step size.

[0136] In step 4206, the larger one of the first consumption and the second consumption is determined as the target consumption in the subsequent time window, that is, target_cost=max(target_cost_1, target_cost_2), where target_cost is the target consumption in the subsequent time window.

[0137] In step 4207, the target consumption in the subsequent time window is determined based on the distribution of expected cumulative consumption, expected single advertisement cost, and actual single advertisement cost. In some embodiments, the specific calculation method may be the same as that in step 4205, that is,

[0138] target_cost = target_cost_2.

[0139] In step 4208, it is determined whether there is actual consumption data for a historical date, that is, whether the advertising unit is a new advertising unit on the day. If it is not a new advertising unit, then there is actual consumption data for a historical date. If there is actual consumption data for a historical date, then step 4209 is executed; otherwise, step 4211 is executed.

[0140] In step 4209, the weighted sum of the actual cumulative consumption of the previous time window of the day and the actual cumulative consumption at the corresponding time of the previous day is taken as the expected consumption value, and then step 4210 is executed. In some embodiments, the weighted sum of the accumulated actual consumption at the corresponding time of the previous time window of the previous day and the accumulated actual consumption at the corresponding time of the previous time window of the day can be used to obtain the expected consumption value, such as calculating the expected consumption value according to the following formula:

[0141] reasonable_cost=cost_base *A + cost*B

[0142] Among them, reasonable_cost is the expected consumption value, A is a hyperparameter that can be set according to the actual scenario, A+B=1, cost_base is the cumulative actual consumption at the corresponding moment of the previous time window on the previous day, and cost is the cumulative actual consumption at the corresponding moment of the previous time window on the current day.

[0143] In step 4210, the target consumption in the subsequent time window is determined based on the expected consumption value, the actual cumulative consumption ratio of the day, and the expected cumulative consumption ratio of the subsequent time window.

[0144] In some embodiments, the target consumption in the subsequent time window may be calculated according to the following formula:

[0145] target_cost = reasonable_cost / ratio_3 * ratio_4

[0146] Among them, ratio_1 is the actual cumulative consumption ratio in the first time window, and ratio_2 is the expected cumulative consumption ratio in the second time window.

[0147] In step 4211, the product of the actual cumulative consumption ratio of the current day and the expected full-day consumption based on the expected single advertisement cost is taken as the target consumption.

[0148] In some embodiments, the target consumption in the subsequent time window may be calculated according to the following formula:

[0149] target_cost = A1*tcpa * ratio_5

[0150] Among them, ratio_5 is the actual cumulative consumption ratio of the first time window, and A1 is a hyperparameter that can be set according to the actual scenario.

[0151] Through this method, the target consumption in the subsequent time window can be dynamically determined as the threshold value for preventing explosion-prone consumption. When the unit consumption reaches or exceeds this value, the unit will be stopped from continuing to play, preventing explosion problems and reducing the pressure on the network and platform. As input data for the layered competition probability control of advertising units, it can help output the competition probability, so that the advertising unit can achieve stable consumption that conforms to real-time changes when playing normally, thereby improving traffic stability.

[0152] In some embodiments, after the target_cost is calculated according to the actual situation of the ad unit, the calculation result is scaled to adjust the strictness of the consumption control, and the target consumption in the subsequent time window is corrected according to the following formula:

[0153] target_cost_final = target_cost * A2

[0154] Among them, target_cost_final represents the target cost of correcting the later time window, and A2 is a hyperparameter, which represents the relaxation coefficient and can be set according to the actual scenario.

[0155] In some embodiments, the time window length can be set to a fixed hour. After obtaining the target_cost_final at a certain time point (hour+1), the target consumption of all hours from (hour+1) to 23:00 on the same day is cyclically updated. Therefore,

[0156] target_cost_hour[j] = target_cost_final / ratio_h1 * ratio_h[j]

[0157] Among them, j belongs to [hour+2, 23], target_cost_hour[j] represents the target consumption after the j-th point update, ratio_h1 is the expected consumption ratio of the (hour+1)th hour, and ratio_h[j] is the expected consumption ratio of the jth hour.

[0158] At this point, the hourly consumption control value has been updated. According to actual conditions, the hourly granularity can be further refined and shortened to minute-level updates or even real-time updates.

[0159] In this way, the target consumption for each time period can be determined step by step, so as to provide rich basic data for subsequent competition adjustments and improve the flexibility of subsequent competition adjustments.

[0160] In some embodiments, in the above step 312, a flowchart of some embodiments of the expected cumulative consumption determination is as follows: Figure 5 shown.

[0161] In step 5121, the value to be consumed on the current date is determined based on the actual cumulative consumption ratio on the current date and the target total consumption value on the current date.

[0162] In step 5122, based on the distribution of actual consumption of historical dates, the value to be consumed on the current date is distributed in the remaining period of the current date to determine the expected cumulative consumption in the subsequent time window. In some embodiments, the hourly consumption distribution ratio of each granularity historical y days can be counted, where y is a hyperparameter representing the number of historical days for statistics, which can be set according to the actual scenario.

[0163] Through this method, the consumption distribution under ideal conditions can be determined based on the distribution of actual consumption in historical data, and then based on this, the consumption distribution of subsequent event windows can be dynamically adjusted according to the actual consumption incurred on that day, thereby improving the adaptability of expected cumulative consumption changes.

[0164] In some embodiments, in the above step 311, the flowchart of some embodiments of determining the target total consumption value is as follows: Figure 6 shown.

[0165] In step 6101, it is determined whether the number of historical dates of the current ad unit is greater than 1, that is, whether it is not a new ad unit at the initial time of the previous day. If it is greater than 1, it is not a new ad unit of the previous day, and step 6102 is executed; otherwise, step 6109 is executed.

[0166] In step 6102, a historical date to be processed is selected for subsequent processing. In some embodiments, a smaller number can be selected as the number of historical dates to be processed based on the number of historical dates and the predetermined maximum number, and the data of historical dates can be retrieved sequentially from the day before the current date.

[0167] In step 6103, it is determined whether the actual conversion value of the currently processed historical date is greater than or equal to a predetermined second conversion value (for example, the predetermined second conversion value is 3). If it is greater than or equal to the predetermined second conversion value, step 6105 is executed; otherwise, step 6104 is executed.

[0168] In step 6104, the actual total consumption value cost corresponding to the historical date is determined as the first total consumption value target_cost1.

[0169] In step 6105, the actual total consumption value of the corresponding historical date is adjusted according to the ratio of the difference between the expected single advertisement cost and the actual single advertisement cost to the expected single advertisement cost, and the first total consumption value is obtained.

[0170] In some embodiments, the first total consumption value may be calculated according to the following formula:

[0171] target_cost1=cost / (1 + w1 * math.tanh(cpa_diff))

[0172] Among them, target_cost1 is the first total consumption, w1 is the first weight, and math.tanh(cpa_diff) represents the hyperbolic tangent function operation on cpa_diff to achieve nonlinear estimation of target_cost1.

[0173] Since the hyperbolic tangent function has the characteristics of simple calculation form, fast calculation speed, symmetry of the origin, and the gradient gradually decreases from the origin, and the gentle part on the left and right is the state of gradient saturation, the cpa_diff can be adjusted consistently in positive and negative cases without preference for positive or negative. When the absolute value of cpa_diff is small, the denominator is large and the adjustment amplitude is small, and when the absolute value of cpa_diff is large, the denominator is small and the adjustment amplitude is large, thereby unifying the goal of the entire adjustment process: that is, adjusting towards the direction of cpa_diff=0.

[0174] In step 6106, the first total consumption value and the expected single advertisement cost are multiplied by the corresponding hyperparameters, and compared with the predetermined minimum target to determine the largest one as the second total consumption value of the current date, for example, according to the following formula:

[0175] target_cost2 = max(target_cost1*w2, w3*tcpa, w4)

[0176] Among them, target_cost2 is the second total consumption value of the current date, w2 and w3 are the weights of the first total consumption value and the expected single advertisement cost respectively, and w4 is the predetermined minimum target.

[0177] In step 6107, it is determined whether there are any historical dates to be processed. If so, step 6102 is executed; otherwise, step 6108 is executed.

[0178] In step 6108, the weighted sum of the second total consumption values ​​of different historical dates is obtained as the target total consumption value target_cost0 of the current date. For example, the weight of the historical date closer to the current date can be set to be larger, or the weight can be reduced or increased for special dates, thereby adjusting the reference value of different dates. In some embodiments, the weight of different historical dates is a hyperparameter, which can be set and adjusted according to the actual scenario.

[0179] In step 6109, it is determined whether the actual conversion value of the currently processed historical date is greater than or equal to the predetermined second conversion value. The currently processed historical date is the only existing historical date. If it is greater than or equal to the predetermined second conversion value, step 6111 is executed; otherwise, step 6110 is executed.

[0180] In step 6110, the actual total consumption value corresponding to the historical date is determined to be the first total consumption value, that is, target_cost1=cost.

[0181] In step 6111, the actual total consumption value of the corresponding historical date is adjusted according to the ratio of the difference between the expected single advertisement cost and the actual single advertisement cost to the expected single advertisement cost, and the first total consumption value is obtained, that is,

[0182] target_cost1=cost / (1 + w5* math.tanh(cpa_diff))

[0183] Among them, w5 is a hyperparameter and can be the same as w1.

[0184] In step 6112, the first total consumption value and the expected single advertisement cost are multiplied by the corresponding hyperparameters respectively, and compared with the predetermined minimum target, and the largest one is determined to be the second total consumption value of the current date, that is,

[0185] target_cost2 = max(target_cost1*w6, w7*tcpa, w8)

[0186] Among them, w6 and w7 are weights of the first total consumption value and the expected single advertisement cost respectively, w8 is the predetermined minimum target, w6 can be the same as w2, w7 can be the same as w3, and w8 can be the same as w4.

[0187] In step 6113, the second total consumption value is used as the target total consumption value for the current date, that is, target_cost0=target_cost2.

[0188] Through this method, it is possible to control consumption on a daily basis based on historical data, dynamically determine the target total consumption, improve the degree of adaptability, and make it easier for competitors to adjust the effect of maintaining traffic smoothness.

[0189] Schematic diagrams of some embodiments of the competition control system disclosed in the present invention are as follows Figure 7 shown.

[0190] The estimation unit 731 can obtain the real-time estimated conversion parameters of the advertisement through the estimation model. In some embodiments, the estimation model can be constructed to predict the advertisement effect of the current request in real time as the real-time estimated conversion parameters.

[0191] The stratification unit 732 can determine the traffic quality layer to which the real-time consumption at the time corresponding to the real-time estimated conversion parameter belongs according to the real-time estimated conversion parameter and the estimated conversion parameter stratification threshold.

[0192] The probability adjustment unit 733 can adjust the contention probability of each traffic quality layer for each traffic quality layer according to the difference between the actual consumption in the previous time window and the target consumption in the subsequent time window in two adjacent time windows.

[0193] The competition unit 734 can conduct advertisement competition according to the adjusted competition probability.

[0194] Such a competition control system can process actual consumption in layers according to real-time estimated conversion parameters, and adjust the competition probability of each traffic quality layer based on the consumption of different layers and target consumption, with the time window as the time processing unit, thereby improving the differentiated processing capabilities of different quality traffic, and can also intervene in the probability of subsequent consumption in a timely manner. On the basis of not reducing or even improving the advertising effect required by advertisers, it can avoid the generation of sudden traffic and improve the stability of competing traffic, while helping to ensure user access speed and user experience.

[0195] In some embodiments, Figure 7As shown, a blocking determination unit 735 can also be included to determine whether data flow blocking occurs. If data flow blocking does not occur, the probability adjustment unit 733 is activated to perform its function; if data flow blocking occurs, a flow blocking alarm is issued so that emergency measures can be automatically or manually executed. In some embodiments, the probability of participating in the competition can be kept unchanged, and the system can exit directly and send an alarm.

[0196] Such a system can promptly detect and handle data flow blockages, improve the operational reliability of advertising competitions, avoid erroneous adjustments to competition probabilities caused by data flow blockages, and improve the reliability of competition control.

[0197] In some embodiments, Figure 7 As shown, a first processing unit 736 may also be included, which can perform corresponding processing for some special situations.

[0198] In some embodiments, the first processing unit 736 can initialize the competition probability of all traffic quality layers of all advertising units to a predetermined first initial value at the beginning of each day. For example, when initialized at 0:00 every day, the competition probability of all advertising units is initialized to the hyperparameter G1, thereby avoiding data deviation caused by daily processing and ensuring operational stability.

[0199] In some embodiments, the first processing unit 736 can initialize the competition probability of all traffic quality layers of the newly added advertising units to a predetermined second initial value when there are newly added advertising units. For example, the competition probability of each layer of the newly added advertising units on that day is initialized to the hyperparameter G2, and the subsequent first adjustment is made based on G2, thereby realizing automatic operation when new advertising units are added.

[0200] In some embodiments, the first processing unit 736 can set the competition probability of the traffic quality layer with the highest threshold value of the estimated conversion interval to the predetermined first probability, and set the competition probability of other traffic quality layers to be less than the predetermined first probability when the target consumption of the subsequent time window is less than or equal to the predetermined second consumption threshold. For example, if the target consumption of the next window is less than or equal to 0, the competition probability is set to 0.01 for the highest layer and 0 for other layers. Through such a method, while reasonably avoiding overspending, it is possible to avoid the system from falling into a state where the competition probability of each layer is 0, thereby improving the robustness of the system.

[0201] In some embodiments, after each adjustment of the competition probability of each traffic quality layer, the first processing unit 736 can adjust the competition probability of the layer with the highest estimated conversion interval threshold corresponding to each traffic quality layer with a competition probability of 0 to a predetermined second probability, for example, setting the highest PCVR layer with a pacing rate of 0 to 0.1, for accumulating data.

[0202] In some embodiments, the first processing unit 736 can modify the actual consumption of the previous time window to the predetermined third consumption if the consumption error does not belong to the predetermined first interval when the actual consumption of the previous time window is less than or equal to the predetermined second consumption threshold. For example, if the previous window consumption of a certain layer is 0, and the competition probability of the layer needs to be adjusted (the consumption error is not within the predetermined first interval), the previous window consumption is set to the default hyperparameter F. Such a method can avoid the situation where the subsequent competition probability changes too much when the actual consumption of the previous window is too small, and further improve the smoothness of the traffic.

[0203] In some embodiments, Figure 7 As shown, it can also include a stratified threshold determination unit 701, which can divide the estimated conversion parameters into a predetermined number of equal parts according to the distribution of the estimated conversion parameters of historical dates of a predetermined number of days and a predetermined number of layers, and obtain the quantile parameters of the divided traffic quality layers as the stratified thresholds of the estimated conversion parameters.

[0204] Such a system can ensure that the stratification of traffic quality layers can match the actual conversion of advertising units, and improve the accuracy and response efficiency of traffic smoothness control.

[0205] In some embodiments, Figure 7 As shown, the window target consumption determination unit 721 may also be included, which can determine the target consumption of the subsequent time window and serve as the calculation basic data of the probability adjustment unit 733. In some embodiments, the window target consumption determination unit 721 can calculate the target consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost.

[0206] Such a system can dynamically determine the target consumption in the subsequent time window based on historical data and real-time actual consumption, improve the reliability of the determined target consumption in the subsequent time window, and further improve the robustness of the flow smoothness control effect.

[0207] In some embodiments, Figure 7 As shown, it can also include an expected cumulative consumption determination unit 712, which can determine the expected cumulative consumption in the subsequent time window according to the distribution of actual consumption on historical dates, the actual cumulative consumption ratio of the current date and the total target consumption on the current date, so as to serve as the data basis for the operation of the window target consumption determination unit 721.

[0208] Such a system can dynamically determine the expected cumulative consumption based on historical data and actual consumption, improve the reliability of the expected cumulative consumption, and further improve the robustness of the flow smoothness control effect.

[0209] In some embodiments, Figure 7 As shown, a target total consumption value determination unit 711 can also be included, which can determine the target total consumption value of the current date based on the actual total consumption value of the historical date, the expected single advertisement cost and the actual single advertisement cost, so as to serve as the data basis for the operation of the expected cumulative consumption determination unit 712.

[0210] Such a system can dynamically determine the target total consumption value for the current date, improve the reliability of the target total consumption value for the current date, and further improve the robustness of the effect of flow smoothness control.

[0211] The structural diagram of an embodiment of the competition control system disclosed in the present invention is as follows Figure 8 As shown. The competition control system includes a memory 801 and a processor 802. Among them: the memory 801 can be a disk, a flash memory or any other non-volatile storage medium. The memory is used to store the instructions in the corresponding embodiment of the competition control method mentioned above. The processor 802 is coupled to the memory 801 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 802 is used to execute the instructions stored in the memory, which can improve the smoothness of the competition flow.

[0212] In one embodiment, it is also possible to Fig. 9 As shown, the competition control system 900 includes a memory 901 and a processor 902. The processor 902 is coupled to the memory 901 via a BUS 903. The competition control system 900 can also be connected to an external storage device 905 via a storage interface 904 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 906. No further detailed description will be given here.

[0213] In this embodiment, the data instructions are stored in the memory, and the instructions are processed by the processor, so that the stability of the competing traffic can be improved.

[0214] In another embodiment, a computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the method in the corresponding embodiment of the competition control method. It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0215] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0216] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0217] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0218] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0219] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0220] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present disclosure rather than to limit it. Although the present disclosure has been described in detail with reference to the preferred embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present disclosure can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present disclosure, which should be included in the scope of the technical solution for protection of the present disclosure.

Claims

1. A competition control method, comprising: Obtain the real-time estimated conversion parameters of the advertisement through the estimation model; Determine, according to the real-time estimated conversion parameter and the estimated conversion parameter stratification threshold, the flow quality tier to which the real-time consumption at the moment corresponding to the real-time estimated conversion parameter belongs; For each traffic quality layer, according to the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows, adjusting the probability of each traffic quality layer participating in the competition; Conducting advertisement bidding according to the adjusted bidding probability, wherein the consumption is the budget consumption for advertisement bidding; The method further includes: determining a target consumption of the subsequent time window according to an expected consumption value, a proportion of actual cumulative consumption of the previous time window, and a proportion of expected cumulative consumption of the subsequent time window, or according to a distribution of expected cumulative consumption, an expected single advertisement cost, and an actual single advertisement cost, wherein the expected consumption value is a weighted sum of the actual cumulative consumption of the previous time window of the day and the actual cumulative consumption at a corresponding moment of the previous day; Before determining to adjust the contention probability of each of the traffic quality layers, determining whether data flow congestion occurs according to the change in the accumulated value of actual consumption; When the difference between the latest accumulated value of actual consumption and the accumulated value of the previous predetermined number of actual consumption is less than or equal to the blocking threshold, it is determined that data flow blocking occurs; Otherwise, the operation of adjusting the contention probability of each traffic quality layer according to the difference between the actual consumption in the earlier time window and the target consumption in the later time window in two adjacent time windows for each traffic quality layer is performed.

2. The method according to claim 1, wherein: The adjusting the contention probability of each of the traffic quality layers comprises: Obtaining a consumption error between actual consumption in an earlier time window and target consumption in a later time window in two adjacent time windows; When the consumption error belongs to a predetermined first interval, the competition probability is not adjusted; When the consumption error does not belong to the predetermined first interval, the contention probability of each traffic quality layer is adjusted in sequence according to the predetermined adjustment strategy in accordance with the order of the estimated conversion interval thresholds corresponding to the traffic quality layers.

3. The method according to claim 2, wherein: The step of obtaining the consumption error between the actual consumption in the previous time window and the target consumption in the subsequent time window of two adjacent time windows includes: According to the time length ratio of the subsequent time window to the previous time window, adjusting the actual consumption of the previous time window to obtain the proportional change consumption; The difference between the target consumption in the subsequent time window and the proportional change consumption is obtained as the consumption error.

4. The method according to claim 2, wherein: When the consumption error does not belong to the predetermined first interval, adjusting the contention probability of each traffic quality layer in sequence according to the predetermined adjustment strategy according to the order of the estimated conversion interval thresholds corresponding to the traffic quality layers includes: When the consumption error is less than a lower threshold of a predetermined first interval, the probability of participating in the competition of each traffic quality layer is reduced in order from low to high according to the estimated conversion interval thresholds corresponding to the traffic quality layers; and When the consumption error is greater than a high threshold of a predetermined first interval, the probability of participating in the competition of each traffic quality layer is increased in sequence according to the estimated conversion interval thresholds corresponding to the traffic quality layers from high to low.

5. The method according to any one of claims 2 to 4, wherein: The step of adjusting the contention probability of each traffic quality layer according to a predetermined adjustment strategy includes: For each traffic quality layer, Obtaining the sum of the consumption error and the actual consumption of the previous time window corresponding to the flow quality layer to determine the adjusted consumption; According to the ratio of the adjusted consumption to the actual consumption of the previous time window of the corresponding traffic quality layer, the contention probability of the previous time window is adjusted to obtain the contention probability of the subsequent time window.

6. The method according to claim 1, further comprising: When the actual consumption of a traffic quality layer is less than a predetermined first consumption threshold, the advertisement bid is increased by a predetermined percentage, and the competition probability of each traffic quality layer is updated until the actual consumption of each traffic quality layer is not less than the predetermined first consumption threshold.

7. The method according to claim 1, wherein: The determining of the target consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost includes: When the actual conversion value is greater than the predetermined first conversion value, if the current actual single advertisement cost on that day is not lower than the expected single advertisement cost, the target consumption in the subsequent time window is determined based on the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost.

8. The method according to claim 7, wherein: The determining of the target consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost, further includes: In the case where the actual conversion value is greater than the predetermined first conversion value, if the actual single advertisement cost on that day is lower than the expected single advertisement cost, then Determine the expected consumption value according to the expected single advertisement cost and the actual conversion value, and determine the first consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window, and the expected cumulative consumption ratio of the subsequent time window; Determining the second consumption in the subsequent time window according to the expected consumption distribution, the expected single advertisement cost and the actual single advertisement cost; The larger one of the first consumption and the second consumption is determined as the target consumption in the subsequent time window.

9. The method according to claim 1, wherein: The determining of the target consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost, further includes: If the actual conversion value is not greater than the predetermined first conversion value, if there is actual consumption data for the historical date, then The weighted sum of the actual accumulated consumption in the previous time window of the day and the actual accumulated consumption at the corresponding time of the previous day is taken as the expected consumption value; The target consumption in the subsequent time window is determined according to the expected consumption value, the proportion of the actual cumulative consumption on the day and the proportion of the expected cumulative consumption in the subsequent time window.

10. The method according to claim 1, further comprising: When the actual conversion value is not greater than the predetermined first conversion value, if there is no actual consumption data for the historical date, the product of the current actual cumulative consumption ratio on the day and the expected full-day consumption based on the expected single advertising cost is used as the target consumption.

11. The method according to claim 1, further comprising: The expected cumulative consumption in the subsequent time window is determined based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio on the current date, and the target total consumption on the current date.

12. The method according to claim 11, further comprising: The target total consumption value for the current date is determined based on the actual total consumption value on historical dates, the expected cost per advertisement, and the actual cost per advertisement.

13. The method according to claim 11, wherein: The determining the expected cumulative consumption in the subsequent time window includes: Determine the value to be consumed on the current date based on the actual cumulative consumption ratio on the current date and the target total consumption value on the current date; According to the distribution of actual consumption of the historical dates, the value to be consumed on the current date is distributed in the remaining time period of the current date to determine the expected cumulative consumption in the subsequent time window.

14. The method according to claim 12, wherein: Determining the target total consumption value includes: When the actual conversion value of the historical date is greater than or equal to the predetermined second conversion value, the actual total consumption value of the corresponding historical date is adjusted according to the ratio of the difference between the expected single advertisement cost and the actual single advertisement cost to the expected single advertisement cost to obtain the first total consumption value; when the actual conversion value of the corresponding historical date is less than the predetermined second conversion value, the actual total consumption value of the corresponding historical date is determined to be the first total consumption value; Multiplying the first total consumption value and the expected single advertisement cost by corresponding hyperparameters respectively, and comparing them with a predetermined minimum target, to determine the largest one as the second total consumption value of the current date; When the number of historical dates is greater than 1, a weighted sum of the second total consumption values ​​of different historical dates is obtained as the target total consumption value of the current date.

15. The method according to claim 14, wherein determining the target total consumption value according to the actual total consumption value of the historical date, the expected single advertisement cost and the actual single advertisement cost further comprises: When the number of historical dates is equal to 1, the second total consumption value is used as the target total consumption value of the current date.

16. The method according to claim 1, further comprising: According to the distribution of the estimated conversion parameters of historical dates of a predetermined number of days and a predetermined number of strata, the estimated conversion parameters are divided into equal parts of the predetermined number of strata, and the quantile parameters for dividing the traffic quality strata are obtained as the estimated conversion parameter stratification thresholds.

17. The method according to claim 1, further comprising at least one of the following: Initializing the competition probabilities of all traffic quality layers of all advertising units to a predetermined first initial value at the beginning of each day; In the case where there is a newly added advertising unit, initializing the competition probability of all traffic quality layers of the newly added advertising unit to a predetermined second initial value; When the target consumption in the subsequent time window is less than or equal to the predetermined second consumption threshold, the competition probability of the traffic quality layer with the highest threshold value in the estimated conversion interval is set to a predetermined first probability, and the competition probabilities of other traffic quality layers are set to be less than the predetermined first probability; or After the adjustment of the contention probability of each of the traffic quality layers is performed each time, the contention probability of the layer with the highest estimated conversion interval threshold corresponding to each of the traffic quality layers with the contention probability of 0 is adjusted to a predetermined second probability.

18. The method according to claim 2, further comprising: In the case where the actual consumption in the previous time window is less than or equal to the predetermined second consumption threshold, if the consumption error does not belong to the predetermined first interval, the actual consumption in the previous time window is modified to the predetermined third consumption.

19. The method according to claim 1, wherein: The upper limit of the competition probability is 1, and the lower limit is 0.

20. A competition control system, comprising: An estimation unit, configured to obtain a real-time estimated conversion parameter of the advertisement through an estimation model; A stratification unit, configured to determine the flow quality stratum to which the real-time consumption at the moment corresponding to the real-time estimated conversion parameter belongs according to the real-time estimated conversion parameter and the estimated conversion parameter stratification threshold; A probability adjustment unit is configured to adjust the competition probability of each traffic quality layer according to the difference between the actual consumption in the previous time window and the target consumption in the subsequent time window in two adjacent time windows for each traffic quality layer; A bidding unit, configured to conduct advertisement bidding according to the adjusted bidding probability, wherein the consumption is a budget consumption for advertisement bidding; The window target consumption determination unit is configured to determine the target consumption of the subsequent time window according to the expected consumption value, the actual cumulative consumption ratio of the previous time window and the expected cumulative consumption ratio of the subsequent time window, or according to the distribution of the expected cumulative consumption, the expected single advertisement cost and the actual single advertisement cost, wherein the expected consumption value is the weighted sum of the actual cumulative consumption of the previous time window of the day and the actual cumulative consumption at the corresponding moment of the previous day; The blocking determination unit is configured to: Before determining to adjust the contention probability of each of the traffic quality layers, determining whether data flow congestion occurs according to the change in the accumulated value of actual consumption; When the difference between the latest accumulated value of actual consumption and the accumulated value of the previous predetermined number of actual consumption is less than or equal to the blocking threshold, it is determined that data flow blocking occurs; Otherwise, the probability adjustment unit is activated.

21. The system of claim 20, further comprising: The expected cumulative consumption determination unit is configured to determine the expected cumulative consumption of the subsequent time window based on the distribution of actual consumption on historical dates, the actual cumulative consumption ratio of the current date and the total target consumption of the current date.

22. The system of claim 21, further comprising: The target consumption total value determination unit is configured to determine the target consumption total value of the current date based on the actual consumption total value of the historical date, the expected single advertisement cost and the actual single advertisement cost.

23. The system of claim 20, further comprising: The stratified threshold determination unit is configured to divide the estimated conversion parameter into a predetermined number of equal parts according to the distribution of the estimated conversion parameter on historical dates of a predetermined number of days and a predetermined number of layers, and obtain the quantile parameters for dividing the traffic quality layers as the estimated conversion parameter stratified thresholds.

24. The system of claim 20, further comprising a first processing unit further configured to perform at least one of the following: Initializing the competition probabilities of all traffic quality layers of all advertising units to a predetermined first initial value at the beginning of each day; In the case where there is a newly added advertising unit, initializing the competition probability of all traffic quality layers of the newly added advertising unit to a predetermined second initial value; When the target consumption in the subsequent time window is less than or equal to the predetermined second consumption threshold, the competition probability of the traffic quality layer with the highest threshold value in the estimated conversion interval is set to a predetermined first probability, and the competition probabilities of other traffic quality layers are set to be less than the predetermined first probability; or After the adjustment of the contention probability of each of the traffic quality layers is performed each time, the contention probability of the layer with the highest estimated conversion interval threshold corresponding to each of the traffic quality layers with the contention probability of 0 is adjusted to a predetermined second probability.

25. A competition control system, comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the method according to any one of claims 1 to 19 based on instructions stored in the memory.

26. A computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 19.

Images