A method, related device, equipment, and storage medium for advertising placement

Abstract

The present application discloses an advertising placement method applied to the field of artificial intelligence. This method can also be implemented using cloud technology to achieve the following steps: obtaining the information on the number of return conversions of P advertisements during a historical time period and the target advertisement during the historical time period; determining the prior distribution parameters of the prior distribution according to the return ratio information; based on the prior distribution parameters and the return conversion number information, determining the target return ratio of the target advertisement in the target time period through the posterior distribution; determining the estimated conversion number of the target advertisement according to the target return ratio and the return conversion number information; determining the sorting result of the target advertisement among the P advertisements according to the estimated conversion number; if it is determined according to the sorting result that the target advertisement meets the advertising placement conditions, then placing the target advertisement. The present application also provides related devices, equipment, and storage media. The present application can calculate a more accurate return ratio, improve the accuracy of the estimated conversion number, increase the accuracy of advertisement sorting, and improve the advertising placement effect.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to an advertisement delivery method, related devices, equipment, and storage medium. Background Art

[0002] With the rapid development of digital media technology, various forms of advertising can reach users through digital media. For enterprises, advertising can reach potential users and obtain immediate or future benefits. Therefore, enterprises continue to increase their investment in advertising, but how to better convert the investment into cash is a question worth exploring.

[0003] Before advertising, we often need to pay attention to the cost of advertising. Currently, in the process of optimizing advertising ranking, we can determine the estimated number of conversions based on the number of return conversions and the return ratio, and then determine the cost adjustment coefficient based on the estimated number of conversions. Then, we can sort the advertisements by the cost adjustment coefficient, and finally give priority to the advertisements with higher rankings.

[0004] The return flow ratio may be different in different time periods, so the return flow ratio can only be estimated based on the statistics of a period of time. However, for low-frequency ads, the corresponding conversion number is small, that is, the fine-grained information is too sparse, which does not conform to the law of large numbers. As a result, there is a large deviation between the estimated return flow ratio and the actual return flow ratio, resulting in inaccurate ad ranking and reduced advertising effectiveness. Summary of the invention

[0005] The embodiments of the present application provide a method, related apparatus, device and storage medium for advertising delivery, which fully consider the dependency between coarse-grained information and fine-grained information. Even if the fine-grained information is sparse, a target return flow ratio that is closer to the actual situation can be calculated, thereby improving the accuracy of the estimated conversion number, which is beneficial to increasing the accuracy of advertising sorting and improving the effectiveness of advertising delivery.

[0006] In view of this, the present application provides, on one hand, a method for delivering advertisements, comprising:

[0007] Obtaining information on the return flow ratios corresponding to P advertisements in a historical time period, and information on the return flow conversion counts corresponding to a target advertisement in a historical time period, wherein the P advertisements at least include the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1;

[0008] Determine a prior distribution parameter corresponding to the prior distribution according to the reflux ratio information;

[0009] Based on the prior distribution parameters and the information of the number of return conversions, determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution, where the target time period belongs to a time period within the historical time period;

[0010] Determine the estimated number of conversions of the target advertisement according to the target return ratio and the information of the number of return conversions;

[0011] Determine the sorting result of the target advertisement among the P advertisements according to the estimated number of conversions;

[0012] If it is determined according to the sorting result that the target advertisement meets the advertisement placement condition, then place the target advertisement.

[0013] On the other hand, the present application provides an advertisement placement device, including:

[0014] An acquisition module, configured to acquire the return ratio information corresponding to the P advertisements in the historical time period, and the information of the number of return conversions corresponding to the target advertisement in the historical time period, where the P advertisements include at least the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1;

[0015] A determination module, configured to determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information;

[0016] The determination module is further configured to determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution based on the prior distribution parameters and the information of the number of return conversions, where the target time period belongs to a time period within the historical time period;

[0017] The determination module is further configured to determine the estimated number of conversions of the target advertisement according to the target return ratio and the information of the number of return conversions;

[0018] The determination module is further configured to determine the sorting result of the target advertisement among the P advertisements according to the estimated number of conversions;

[0019] A placement module, configured to place the target advertisement if it is determined according to the sorting result that the target advertisement meets the advertisement placement condition.

[0020] In a possible design, in an implementation manner of the other aspect of the embodiments of the present application,

[0021] The acquisition module is specifically configured to acquire the return ratio information corresponding to the P advertisements in the historical time period, where the historical time period includes N time periods, the return ratio information includes N return ratios, and the return ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1;

[0022] Obtain the information on the number of return conversions corresponding to the target advertisement in the historical time period, where the information on the number of return conversions includes N numbers of return conversions, and there is a corresponding relationship between the number of return conversions and the time period, and N numbers of return conversions;

[0023] Among them, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement. A low-frequency advertisement is an advertisement whose sum of N numbers of return conversions is less than or equal to the return conversion threshold, and a non-low-frequency advertisement is an advertisement whose sum of N numbers of return conversions is greater than the return conversion threshold.

[0024] In a possible design, in an implementation manner on the other hand of the embodiment of the present application, the prior distribution is the first beta distribution;

[0025] The determination module is specifically configured to calculate the average value and variance according to the return ratio information;

[0026] Determine the first prior parameter in the prior distribution parameters according to the average value and variance;

[0027] Determine the second prior parameter in the prior distribution parameters according to the average value and variance, where the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

[0028] In a possible design, in an implementation manner on the other hand of the embodiment of the present application,

[0029] The determination module is specifically configured to determine the binomial distribution according to the information on the number of return conversions;

[0030] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution;

[0031] Based on the prior distribution parameters and the information on the number of return conversions, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0032] In a possible design, in an implementation manner on the other hand of the embodiment of the present application,

[0033] The determination module is specifically configured to determine the second beta distribution according to the binomial distribution and the first beta distribution, where the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution;

[0034] Determine the Bayesian estimation formula according to the second beta distribution.

[0035] In a possible design, in an implementation manner on the other hand of the embodiment of the present application,

[0036] A determination module, specifically configured to obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number is one of the return conversion numbers in the return conversion number information;

[0037] Based on the first prior parameter, the second prior parameter, and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0038] In a possible design, in an implementation manner of another aspect of the embodiments of the present application, the prior distribution is the first Dirichlet distribution;

[0039] A determination module, specifically configured to calculate the average value and variance according to the return ratio information;

[0040] Determine N prior parameters in the prior distribution parameters according to the average value and variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

[0041] In a possible design, in an implementation manner of another aspect of the embodiments of the present application,

[0042] A determination module, specifically configured to determine the multinomial distribution according to the return conversion number information;

[0043] According to the multinomial distribution and the first Dirichlet distribution, determine the Bayesian estimation formula corresponding to the posterior distribution;

[0044] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0045] In a possible design, in an implementation manner of another aspect of the embodiments of the present application,

[0046] A determination module, specifically configured to determine the second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution;

[0047] Determine the Bayesian estimation formula according to the second Dirichlet distribution.

[0048] In a possible design, in an implementation manner of another aspect of the embodiments of the present application,

[0049] A determination module, specifically configured to obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number is one of the return conversion numbers in the return conversion number information;

[0050] Based on N prior parameters and the target reflux conversion number, the target reflux ratio corresponding to the target advertisement in the target time period is calculated through the Bayesian estimation formula.

[0051] In a possible design, in an implementation manner of another aspect of the embodiments of the present application,

[0052] A determination module, specifically configured to determine a future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number;

[0053] According to the future cost achievement adjustment coefficient corresponding to the target advertisement, determine the revenue per thousand impressions corresponding to the target advertisement;

[0054] Obtain the revenue per thousand impressions corresponding to each advertisement to be placed among P advertisements, where the revenue per thousand impressions corresponding to each advertisement to be placed is determined according to the future cost achievement adjustment coefficient corresponding to each advertisement to be placed;

[0055] Sort the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each advertisement to be placed, and obtain the sorting result of the target advertisement among the P advertisements.

[0056] In a possible design, in an implementation manner of another aspect of the embodiments of the present application, the advertisement placement device further includes a push module and a reception module;

[0057] The push module is configured to push the sorting result of the target advertisement among the P advertisements to the client;

[0058] The reception module is configured to receive an advertisement placement selection instruction sent by the client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, and Q is an integer greater than or equal to 1 and less than or equal to P;

[0059] The determination module is further configured to determine that the target advertisement meets the advertisement placement condition if the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement.

[0060] Another aspect of the present application provides a computer device, including: a memory, a transceiver, a processor, and a bus system;

[0061] Wherein, the memory is used to store programs;

[0062] The processor is configured to execute the programs in the memory, and the processor is configured to execute the methods described in the above aspects according to the instructions in the program code;

[0063] The bus system is used to connect the memory and the processor to enable the memory and the processor to communicate with each other.

[0064] On the other hand, the present application provides a computer-readable storage medium storing instructions which, when run on a computer, cause the computer to execute the methods described in the above aspects.

[0065] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages:

[0066] The present application provides a method for advertising placement. First, obtain the return ratio information corresponding to P advertisements in a historical time period and the return conversion number information corresponding to a target advertisement in the historical time period. Then, determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information. Next, based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in a target time period through the posterior distribution. According to the target return ratio and the return conversion number information, determine the estimated conversion number of the target advertisement. Finally, determine the sorting result of the target advertisement among the P advertisements according to the estimated conversion number. If it is determined that the target advertisement meets the advertising placement conditions according to the sorting result, then the target advertisement can be placed. In the above manner, the return ratio information statistically obtained by the same advertiser in the historical time period is used as the basis for constructing the prior probability. This return ratio information belongs to coarse-grained information, and the return conversion number information of the target advertisement is used as sample information, which belongs to fine-grained information. Based on Bayes' theorem, the posterior distribution can be derived according to the prior probability and the sample information, and then the target return ratio of the target advertisement can be determined based on the posterior distribution. Thus, introducing coarse-grained information as prior knowledge and fully considering the dependence relationship between coarse-grained information and fine-grained information, even if the fine-grained information is sparse, it is possible to calculate a target return ratio closer to the real situation, thereby improving the accuracy of the estimated conversion number, facilitating an increase in the accuracy of advertisement sorting, and enhancing the advertising placement effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a schematic diagram based on conversion delay in an embodiment of the present application;

[0068] Figure 2 It is a schematic diagram for determining the estimated conversion number based on the reported conversion number and the return ratio in an embodiment of the present application;

[0069] Figure 3 It is a schematic diagram of the relationship between coarse-grained information and fine-grained information in an embodiment of the present application;

[0070] Figure 4 It is a schematic diagram of the relationship between the target cost and the current cost in an embodiment of the present application;

[0071] Figure 5 It is a schematic diagram of an interaction environment of an advertising placement system in an embodiment of the present application;

[0072] Figure 6 It is a schematic diagram of an architecture of an advertising system in an embodiment of the present application;

[0073] Figure 7 It is a schematic diagram of an embodiment of an advertising placement method in an embodiment of the present application;

[0074] Figure 8 It is a schematic diagram of an embodiment of adjusting the sorting of target advertisements in an embodiment of the present application;

[0075] Figure 9 It is a schematic illustration of the corresponding return ratios of N time periods within a historical time period in an embodiment of the present application;

[0076] Figure 10 It is a schematic diagram of a posterior probability density based on the first beta distribution in an embodiment of the present application;

[0077] Figure 11 It is a schematic diagram of an interface for pushing the sorting result of target advertisements in an embodiment of the present application;

[0078] Figure 12 It is a schematic diagram of an interface for triggering an advertising placement selection instruction in an embodiment of the present application;

[0079] Figure 13 It is a schematic diagram of an embodiment of an advertising placement device in an embodiment of the present application;

[0080] Figure 14 It is a schematic diagram of the structure of a computer device in an embodiment of the present application. Detailed implementation manners

[0081] The embodiments of the present application provide a method, related device, equipment, and storage medium for advertising placement, which fully consider the dependency relationship between coarse-grained information and fine-grained information. Even if the fine-grained information is sparse, it can calculate a target return ratio closer to the actual situation, thereby improving the accuracy of estimated conversion numbers, facilitating increasing the accuracy of advertisement sorting, and enhancing the advertising placement effect.

[0082] The terms "first", "second", "third", "fourth", etc. (if any) in the description, claims and the above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented, for example, in an order other than those illustrated or described herein. In addition, the terms "comprising" and "corresponding to" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0083] It should be understood that the present application provides an advertising placement method based on Artificial Intelligence (AI) technology. This method is applicable to online advertising placement on websites. For example, it can be used for advertising placement on search engines, in information flow products, on video websites, and on TV, etc. AI technology aims to provide intelligent marketing strategies for advertisers. How to push advertisements that users are interested in and place the products that users are most likely to purchase (with high conversion rates) at the top position is a key issue. This can not only greatly increase the profits of enterprises but also improve user stickiness to a certain extent. The advertising placement method provided by the present application can accurately recommend advertisements to users, making the advertising placement more precisely reach the target population and achieving more cost-effective and efficient marketing.

[0084] It can be understood that AI uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, including theories, methods, technologies, and application systems for perceiving the environment, acquiring knowledge, and using knowledge to obtain the best results. In other words, AI is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. AI also studies the design principles and implementation methods of various intelligent machines to enable the machines to have functions of perception, reasoning, and decision-making. Among them, AI technology is an interdisciplinary subject involving a wide range of fields, including both hardware-level technologies and software-level technologies. AI basic technologies generally include technologies such as sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. AI software technologies mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0085] The present application involves many professional terms. To better understand the present application, the following will introduce these professional terms separately.

[0086] 1. Optimized Cost per Action (oCPA): After the advertiser selects the optimization goal and makes a bid, conversion effect data is also fed back. After estimation, the bid and the actual conversion cost are balanced. Among them, the optimized actions include but are not limited to activation, registration, and placing an order, etc. Essentially, oCPA is still charged according to the Cost per Action (CPA).

[0087] 2. Optimized Cost per Click (oCPC): By adopting a conversion rate estimation mechanism, it provides high-quality traffic to advertisers and ensures the conversion rate. Based on the advertiser's bid, the system will dynamically adjust the bid based on multi-dimensional, real-time feedback, and historical accumulated data, and according to the estimated conversion rate and competitive environment, so as to optimize the ad ranking, help advertisers find suitable traffic, and reduce the conversion cost. Essentially, oCPC is still charged according to the Cost per Click (CPC).

[0088] 3. Optimized Cost per Mille (oCPM): By adopting a more accurate click-through rate and conversion rate estimation mechanism, the ad is shown to users who are suitable for the ad and are likely to convert, improving the ad's conversion rate and reducing the conversion cost. Essentially, oCPM is still charged according to the Cost per Mille (CPM).

[0089] 4. Real-time price adjustment algorithm: It refers to an algorithm that adjusts the online bid through a cost achievement adjustment coefficient to achieve the purpose of cost control.

[0090] 5. Cost achievement adjustment coefficient: In the online oCPA bidding and ranking formula, there is a cost achievement adjustment coefficient that can adjust the bid to achieve the purpose of cost control.

[0091] 6. Conversion feedback: The advertiser feeds back the conversion to the background. Among them, "conversion" means the behavior of purchase, registration, or information demand affected by the online advertisement.

[0092] 7. Conversion delay: There is usually a long time interval from when a user clicks on an ad to when the ad system learns that the user has activated the application. This is mainly caused by the following two reasons. On the one hand, the user may start the application a long time after downloading it. On the other hand, the user's action of starting the application needs to be reported and fed back to the ad system by the advertiser, so there will be a certain delay. For the convenience of introduction, please refer to Figure 1 , Figure 1This is a schematic diagram based on conversion delay in the embodiments of the present application. As shown in the figure, it usually only takes a few seconds or minutes from ad exposure to ad click, only a few seconds from ad click to downloading the application corresponding to the ad, and a few minutes from ad click to installing the application corresponding to the ad. However, it usually takes several days from ad click to paying in the application corresponding to the ad. Therefore, there is a long time delay in the conversion backflow after the click operation occurs.

[0093] 8. Conversion backflow number: It is a random variable representing the number of conversions that have flowed back.

[0094] 9. Backflow ratio: It represents the ratio of the conversion backflow number to the total conversion number within a certain period. In this application, the total conversion number to be estimated is called the estimated conversion number (conv h ).

[0095] 10. First-day backflow ratio: It represents the ratio of the conversion backflow number on the first day to the total conversion number.

[0096] 11. Estimated conversion number (conv h ): That is, the total conversion number to be estimated. In the process of determining the estimated conversion number using the conversion backflow number and the backflow ratio, a backflow ratio needs to be estimated. For the convenience of introduction, please refer to Figure 2 , Figure 2 This is a schematic diagram based on the conversion backflow number and the backflow ratio to determine the estimated conversion number in the embodiments of the present application. As shown in the figure, the backflow ratio (ratio) is estimated based on the converted and unconverted flows within 15 hours. The estimated conversion number (conv h ) can be calculated in the following way:

[0097]

[0098] However, there is a certain error between the backflow ratio (ratio) and the true backflow ratio . Therefore, there is also a certain error between the estimated conversion number (conv h ) and the actual conversion number .

[0099] 12. Conversion backflow window: Usually, the conversion backflow window is set to 5 days, that is, the conversion backflow time is 5 days, and the conversion backflow numbers greater than 5 days will no longer be used. It should be noted that the size of the conversion backflow window can also be adjusted according to actual needs, that is, the conversion backflow time can be adjusted, which is not limited in this application.

[0100] 13. Coarse-grained information: Usually includes the conversion backflow number and the total conversion number of all ads placed by the same advertiser, which can be expressed as advertiser * product identifier, where "*" represents permutation and combination. For the convenience of explanation,Figure 3 This is a schematic diagram of the relationship between coarse-grained information and fine-grained information in the embodiments of the present application. As shown in the figure, the same advertiser usually has multiple products to be promoted through advertisements. Each product has a corresponding product identifier. For example, the identifier of product A is 1, and the identifier of product B is 2. The relationship between a product and an advertisement can be one-to-one or one-to-many. For example, product A is promoted through advertisement A, and at the same time, product A is also promoted through advertisement B. However, usually one advertisement only corresponds to one product. Figure 3 Advertisement A and advertisement B in [description] both belong to the advertisements placed by the same advertiser. Suppose the advertiser has two products, namely product 1 and product 2. Based on this, the results of advertiser * product identifier include four cases, namely product 1 + advertisement A, product 1 + advertisement B, product 2 + advertisement A, and product 2 + advertisement B. The theoretical return ratios in these four cases should be close.

[0101] 14. Fine-grained information: It represents the return conversion number and total conversion number of a certain advertisement placed by the same advertiser. For the convenience of explanation, please refer to Table 1. Table 1 is a schematic diagram based on the coarse-grained information and fine-grained information corresponding to the same advertiser.

[0102] Table 1

[0103] Number of conversions from first-day return Total number of conversions First-day return ratio Advertisement A 1 2 50％ Advertisement B 2 2 100％ Advertiser * Product identifier 300 1000 30％

[0104] As can be seen from Table 1, for the coarse-grained information, relatively sufficient data can be obtained within the first day. For the fine-grained information (such as advertisement A or advertisement B), the data obtained within the first day is relatively sparse.

[0105] 15. Low-frequency advertisement: It represents an advertisement with a small number of return conversions in the historical time period. Specifically, if within the historical time period, the return conversion number of an advertisement is less than or equal to the return conversion number threshold, then this advertisement is a low-frequency advertisement. It should be noted that the historical time period can be 24 hours, or 5 days, or other time lengths. The return conversion number threshold can be 10, or 15, or other values. In the present application, the historical time period is taken as 24 hours and the return conversion number threshold is taken as 10 as an example for introduction. However, this should not be construed as a limitation of the present application. Since the return conversion number of low-frequency advertisements is small and does not conform to the law of large numbers, the deviation of the estimated return ratio is larger.

[0106] 16. Non-low-frequency advertisement: It refers to an advertisement with a relatively large number of return conversions within a historical time period. Specifically, if within a historical time period, the number of return conversions of a certain advertisement is greater than the return conversion threshold, then this advertisement is a non-low-frequency advertisement. It should be noted that the historical time period can be 24 hours, or 5 days, or other time lengths, and the return conversion threshold can be 10, or 15, or other values. In this application, the historical time period is 24 hours and the return conversion threshold is 10 as an example for introduction, but this should not be construed as a limitation of this application. Since the number of return conversions of non-low-frequency advertisements is relatively large and conforms to the law of large numbers, the estimated return ratio is more accurate than the return ratio estimated based on low-frequency advertisements.

[0107] Based on the above introduction, to better understand the application background of this application, the application background related to this application will be introduced in detail below.

[0108] It should be understood that for the media party, on the one hand, it pursues the long-term experience of the platform product, and on the other hand, it pursues the maximization of traffic benefits. The formula for the total media advertising revenue is:

[0109] Total revenue = total advertising traffic × traffic fill rate × price per exposure;

[0110] Among them, the price per exposure is the effective cost per mile (eCPM). The traffic fill rate is the ratio of the number of advertisement displays (or advertisement exposure numbers) to the number of advertisement display opportunities (or advertisement request numbers) within a period of time. Traffic fill rate = number of advertisement displays / number of advertisement display opportunities × 100%.

[0111] If the advertiser expects to maximize the effect, the most direct optimization strategy is to obtain a sufficient number of good conversions at the lowest traffic price (i.e., eCPM). Taking the CPM billing model, CPC billing model, CPA billing model, and oCPA billing model as examples, the method of calculating eCPM under different billing models will be introduced below.

[0112] Please refer to Table 2, which is a schematic diagram of calculating eCPM under different billing models.

[0113] Table 2

[0114]

[0115] Based on the content shown in Table 2, "bid price" refers to the price that an advertiser pays for each click of an advertisement, "billing" refers to the final deduction method after an advertisement is clicked. Among them, "exposure" means that when a user visits a media website and the media displays an advertisement, it is exposed once every time the user sees it. "Click" means that the user clicks once after the advertisement is exposed. "Conversion" means the activation, purchase, registration and other behaviors of the user affected by the online advertisement. "Target relevance" means that if the advertiser's goal is to obtain conversions and the advertising delivery mechanism is to ensure conversions, then the target relevance is high; if the advertising delivery mechanism is to ensure clicks or other types, then the target relevance is not high. "Conversion cost control" means controlling the cost of conversions. "Revenue stability" means the degree of stability of the revenue obtained by the traffic provider (or media provider).

[0116] The traffic provider (or media provider) can be understood as a website or an application. For example, a certain news application (application, APP) belongs to the traffic provider (or media provider). An advertiser is an enterprise that obtains targeted services. For example, the advertiser of a certain game aims to obtain game users, so it needs to place game advertisements on websites or applications to let users download the game.

[0117] In the CPM billing mode, the revenue of the traffic provider (or media provider) can be guaranteed, while the conversion cost of the advertiser is uncontrollable. In the CPC billing mode, users cannot optimize the conversion cost, there is no way to express their true goals, and there are no effective optimization means. Therefore, it is necessary to estimate the click bid price based on the target conversion cost. Not only can't it bid finely for different traffic, but also the conversion cost is uncontrollable. In the CPA billing mode, the media provider (or traffic provider) bears the revenue risk brought by the pCVR and the prediction error of the predicted click-through rate (Predict Click-Through Rate, pCTR), as well as the risk brought by conversion cheating. In this case, oCPA, oCPC, and oCPM came into being. The following will take the oCPA billing mode as an example for introduction.

[0118] In the bidding ranking formula of the oCPA billing mode, SmartBid = TargetCPA × billing ratio coefficient × cost achievement adjustment coefficient, where TargetCPA represents the cost that the advertiser is willing to pay, and the billing ratio coefficient represents the ratio between the actual deduction after a click and the TargetCPA. Based on this, the eCPM calculation method of the CPM billing mode is as follows:

[0119] eCPM = (TargetCPA × billing ratio coefficient × cost achievement adjustment coefficient) × (pCVR × pCVR correction coefficient) × pCTR;

[0120] Among them, the pCVR correction coefficient corrects pCVR based on historical data, so that pCVR is more accurate. Since the advertiser's goal is to control the conversion cost of the advertisement, therefore, in order to achieve the purpose of cost control, a cost achievement adjustment coefficient is added to the oCPA bidding ranking formula. This cost achievement adjustment coefficient is willing to adjust the bid price, so as to control the consumption and cost of the advertisement.

[0121] For the sake of illustration, please refer to Figure 4 , Figure 4 which is a schematic diagram of the relationship between the target cost and the current cost in the embodiment of the present application. As shown in the figure, the advertiser first sets a target cost (TagetCPA). Assume that the current time is t. The time period from 0 to t can be used as the historical time period, or several days or hours can be used as the historical time period. In the historical time period, known quantities can be obtained. Here, the known quantities include the historical cumulative consumption (cost h ) corresponding to the historical time period, the historical cumulative conversion number (conv h ) corresponding to the historical time period, and the historical cost achievement adjustment coefficient (λ h ) corresponding to the historical time period. And the time period from t to 24 o'clock belongs to the future time period. In the future time period, unknown quantities can be estimated. Here, the unknown quantities include the future cumulative consumption (cost e ) from t to 24 o'clock, the future cumulative conversion number (conv e ) from t to 24 o'clock, and the future cost achievement adjustment coefficient (λ e ).

[0122] For the future cumulative consumption (cost e ), the following method can be used for estimation:

[0123] 1. There is a functional relationship between the consumption and the cost achievement adjustment coefficient (λ), which can be fitted based on data, that is:

[0124]

[0125] Among them, cost h_1.0 represents the consumption corresponding to not adjusting the price in the historical time period. In the case of not adjusting the price, the historical cost achievement adjustment coefficient (λ h ) = 1.

[0126] 2. The proportion of non-price-adjusted consumption over time has a functional relationship and is fitted based on the overall market data, that is:

[0127]

[0128] Among them, cost e_1.0Represents the consumption corresponding to no price adjustment in the future time period.

[0129] 3. Given the future cost achievement adjustment coefficient (λ e ), the future cumulative consumption (cost e ) can be calculated as follows:

[0130]

[0131] Thus, the variable to be solved is obtained, which is a function of the future cost achievement adjustment coefficient (λ e ).

[0132] For the future cumulative conversion number (conv e ), it can be estimated in the following way:

[0133] 1. In the future cost estimation, assume that there is a functional relationship between the average conversion cost (CPA 1.0 ) and the cost achievement adjustment coefficient (λ), and approximately obtain it based on the market data:

[0134]

[0135] Among them, Figure 4 the current cost shown is CPA h , and CPA h is equal to the ratio of the historical cumulative consumption (cost h ) to the historical cumulative conversion number (conv h ). At time t, there is a certain cost deviation between the current cost (CPA h ) and the target cost (TagetCPA), and in the future time period, there may be a large error between the estimated cost and the actual cost. Among them, the estimated cost is equal to the future cumulative consumption (cost e ) divided by the future cumulative conversion number (conv e ).

[0136] 2. In the future conversion estimation, based on the future cumulative consumption (cost e ) and the future cost per action (CPA e ), the following formula can be obtained:

[0137]

[0138] Thus, the variable to be solved is obtained, which is a function of the future cost achievement adjustment coefficient (λ e ).

[0139] Based on this, in the principle of the cost control strategy, the initial problem is to solve the future cost achievement adjustment coefficient (λ e), and the optimization goal is that the final cost is close to the target cost (TargetCPA), that is:

[0140]

[0141] Therefore, the ultimate problem is how to estimate the future cost adjustment coefficient (λ e ). Substituting the future cumulative consumption (cost e ) and the future cumulative conversion number (conv e ) into the optimization goal, the following estimation method can be solved:

[0142]

[0143] As can be seen from the above formula, the target cost (TargetCPA) is ultimately only related to the historical cumulative conversion number (conv h ). That is to say, the prediction of the historical cumulative conversion number (conv h ) is the key. Therefore, in this application, the historical cumulative conversion number (conv h ) is uniformly described as the "predicted conversion number".

[0144] In the conversion data of deep target advertising, it is very sparse, usually only a few conversions. Therefore, if the return ratio is used for prediction, there is still a large inherent deviation. Assuming that the return ratio of each conversion within the window is ratio, then the actual returned conversion number (rc h ) within the time window follows a binomial distribution B(conv h , ratio), the expectation is conv h × ratio, and the variance is conv h × ratio × (1 - ratio), that is:

[0145]

[0146] For easy understanding, please refer to Table 3, which compares the deviation between the actual conversion number and the predicted conversion number (conv h ).

[0147] Table 3

[0148]

[0149] Based on this, the technical solution provided by this application can better estimate the historical cumulative conversion number (conv h ), so as to better control the target cost (TargetCPA).

[0150] This application proposes a method for advertising placement, which is applied toFigure 5 For the advertised delivery system shown, please refer to Figure 5 , Figure 5 which is a schematic diagram of an interaction environment of the advertised delivery system in an embodiment of the present application. As shown in the figure, the advertised delivery system includes a terminal device and a server. The user views advertisements through the terminal device and triggers relevant operations through the terminal device, such as clicking, downloading, installing, activating, placing an order, and paying. The terminal device reports the operation information within a period of time (or instantaneously) to the server. The server performs statistics and calculations based on the collected operation information, determines the ranking of the advertisements according to the calculation results, and then pushes the new ranking results to the terminal device. Thus, the user can view the advertisements with the changed ranking through the terminal device. So far, an update of the advertisement ranking is completed.

[0151] In an online advertisement system, the terminal device immediately feeds back the user's operation information to the server, and the server will adjust the advertisement ranking in real time according to the operation information, so as to achieve the purpose of online adjustment of the advertisement ranking.

[0152] In an offline advertisement system, the server collects the user's operation information within a period of time and then adjusts the advertisement ranking according to the operation information, so as to achieve the purpose of offline adjustment of the advertisement ranking.

[0153] It should be noted that Figure 5 the number and type of the terminal device and the server shown are only for illustration. In actual applications, the terminal device includes but is not limited to tablet computers, laptop computers, personal digital assistants, mobile phones, voice interaction devices, and personal computers (PCs). The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and AI platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, and the present application does not make any restrictions here.

[0154] Due to the extremely large amount of advertising content and advertising quantity, in practical applications, the advertising placement method provided by this application can adopt cloud technology to implement the calculation and sorting of a large number of advertisements. Specifically, cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or a local area network to achieve data calculation, storage, processing, and sharing. Cloud technology is the general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model, which can form a resource pool, be used as needed, and is flexible and convenient. Cloud computing technology will become an important support. The background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites, and more portal websites. Along with the highly developed and applied Internet industry, in the future, each item may have its own identification mark and needs to be transmitted to the background system for logical processing. Data at different levels will be processed separately, and various types of industry data require a powerful system backup support, which can only be achieved through cloud computing.

[0155] Cloud computing is a computing model that distributes computing tasks across a resource pool composed of a large number of computing devices, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides resources is called the "cloud". The resources in the "cloud" seem to be infinitely expandable to users, and can be obtained at any time, used on demand, expanded at any time, and paid according to usage.

[0156] As a basic capability provider of cloud computing, a cloud computing resource pool (referred to as a cloud platform, generally called an Infrastructure as a Service (IaaS) platform) will be established, and various types of virtual resources will be deployed in the resource pool for external customers to choose and use. The cloud computing resource pool mainly includes: computing devices (virtual machines, including operating systems), storage devices, and network devices. According to logical function division, a Platform as a Service (PaaS) layer can be deployed on the IaaS layer, and a Software as a Service (SaaS) layer can be deployed on top of the PaaS layer. It is also possible to directly deploy SaaS on the IaaS. PaaS is the platform for software operation, such as databases, web containers, etc. SaaS are various business software, such as web portal websites, SMS mass senders, etc. Generally speaking, SaaS and PaaS are the upper layers relative to IaaS.

[0157] Furthermore, the advertising placement system can also provide AI cloud services (AI as a Service, AIaaS) for advertisers. This is a mainstream service mode of an AI platform. Specifically, the AIaaS platform splits several common AI services and provides independent or packaged services in the cloud. This service mode is similar to opening an AI-themed mall. All developers can access and use one or more AI services provided by the platform through the Application Programming Interface (API). Some senior developers can also use the AI framework and AI infrastructure provided by the platform to deploy and operate their own exclusive cloud AI services.

[0158] It should be understood that the advertising placement system provided in this application is part of the advertising system. For the convenience of introduction, please refer to Figure 6 , Figure 6 which is an architecture schematic diagram of the advertising system in the embodiment of this application. A typical advertising system is as Figure 6 shown. An advertisement has to go through the processes of request, exposure, rough ranking, and refined ranking. Users can view the exposed advertisement A through a terminal device and can also perform a series of subsequent operations based on the advertisement A (such as click, download, installation, and payment, etc.). These operations enter the advertising placement system through the traffic access layer. The portrait retrieval module in the advertising placement system calls the portrait database to obtain the portrait corresponding to the user. The advertisement retrieval module calls the advertisement database to obtain the advertisement data of advertisement A. The model training module calls the log data in the log library and trains the conversion prediction model based on the advertisement data of advertisement A and the portrait of the user. In addition, the model training module can also predict the eCPM of the advertisement through the conversion prediction model, so as to sort the online advertisements. Before sorting, it needs to go through two processes of rough selection and refined selection. The advertisements after refined selection will be exposed.

[0159] Rough selection mainly uses the lightweight conversion rate (LiteCVR) model included in the conversion prediction model. The formula for sorting using the lightweight conversion rate (LiteCVR) model is as follows:

[0160] eCPM = SmartBid × LiteCVR × LiteCVR × 1000;

[0161] where SmartBid represents the smart bid.

[0162] Refined selection mainly uses the predicted conversion rate (pCVR) model included in the conversion prediction model. The formula for sorting using the predicted conversion rate (pCVR) model is as follows:

[0163] eCPM = SmartBid × pCVR × pCTR × 1000;

[0164] Among them, SmartBid represents the smart bid, and pCTR represents the prediction of the ad click-through rate.

[0165] After selection, the ads are sorted from largest to smallest according to eCPM, and these ads are exposed.

[0166] The solution provided by the embodiments of this application involves technologies such as machine learning of AI. Combining the above introduction, the method for ad delivery in this application will be introduced below. Please refer to Figure 7 , an embodiment of the ad delivery method in the embodiments of this application includes:

[0167] 101. Obtain the return ratio information corresponding to P ads in the historical time period, and the return conversion number information corresponding to the target ad in the historical time period, where the P ads include at least the target ad, and the P ads all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1;

[0168] In this embodiment, taking the target ad as an example for specific illustration, it can be understood that in actual applications, each ad to be delivered can be processed using the ad delivery method provided in this application. The ad delivery device needs to obtain the return ratio information corresponding to P ads belonging to the same advertiser in the historical time period. It should be noted that the historical time period can include multiple time periods. Assuming the historical time period is 24 hours and each time period is one hour, then the historical time period includes 24 time periods. Therefore, the return ratio information consists of 24 return ratios. Also assuming the historical time period is 48 hours and each time period is one hour, then the historical time period includes 48 time periods. Therefore, the return ratio information consists of 48 return ratios, which is not limited here.

[0169] Specifically, in actual applications, the ad delivery device can obtain the return conversion numbers of P ads in the historical time period. Assuming the historical time period includes 24 time periods, then 24 return conversion numbers corresponding to the P ads are obtained. Based on this, when the ad delivery device knows the total conversion number corresponding to each time period, it can calculate the return ratio corresponding to each time period respectively. When obtaining the return ratio corresponding to each time period in the historical time period, the return ratio information corresponding to P ads in the historical time period can be obtained.

[0170] The advertisement placement device also needs to obtain the information on the number of return conversions of the target advertisement within a historical time period, and the target advertisement belongs to one of the P advertisements. Similarly, assuming that the historical time period is 24 hours and each time period is one hour, then the historical time period includes 24 time periods. Therefore, the information on the number of return conversions consists of 24 numbers of return conversions. Also assuming that the historical time period is 48 hours and each time period is one hour, then the historical time period includes 48 time periods. Therefore, the information on the number of return conversions consists of 48 numbers of return conversions, which is not limited here.

[0171] It should be noted that the advertisement placement device is deployed on a computer device, which can be a server or a terminal device with high computing power, and is not limited here.

[0172] The target advertisement includes but is not limited to advertisements in the game industry, website portal industry, e-commerce industry, financial industry, education industry, and tourism industry. Based on the target advertisements in different industries, the reported types of return conversions are often different. For the convenience of understanding, please refer to Table 4, which is a schematic diagram of the types of return conversions based on different industries.

[0173] Table 4

[0174] Industry type Reportable conversion types Games Activation, registration, second-day retention, and payment behavior Website portal Activation, registration, and second-day retention E-commerce Activation, registration, product details page view, search, favorite, add to cart, and order placement Finance Form appointment, web consultation, activation, second-day retention, registration, application, and payment Education Form appointment, web consultation, order placement, and payment Travel Form appointment, web consultation, order placement, and payment

[0175] The content in Table 4 is only for illustration, and the advertiser can also upload the types of return conversions that it is concerned about according to the characteristics of the industry it belongs to.

[0176] 102. Determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information;

[0177] In this embodiment, based on the return ratio information, the advertisement placement device can obtain the prior distribution parameters corresponding to the prior distribution by using the method of moment estimation. This application combines Bayes' theorem to infer the target return ratio of the target advertisement, realizes a probability estimation with higher confidence, and thus makes the estimated number of conversions of the target advertisement more accurate.

[0178] Specifically, when the return ratio information corresponding to the P advertisements within the historical time period is given, determine the best hypothesis in the hypothesis space. For the convenience of description, the return ratio information corresponding to the P advertisements within the historical time period is used as the training data D, and the hypothesis space is set as θ. The "hypothesis space" in this application is the return ratio to be solved. Bayes' theory provides a method for calculating the probability of a hypothesis, that is, based on the training data D, Bayes' theorem is used to update the prior probability P(θ) to obtain the posterior probability P(θ|D). The Bayes' formula is expressed as:

[0179] P(θ|D) ∝ P(D|θ) × P(θ);

[0180] Among them, P(θ) represents the prior probability, P(θ|D) represents the posterior probability of the hypothesis space θ, P(D|θ) represents the likelihood function of the training data D, the symbol "∝" represents proportional to, and the symbol "|" represents conditional.

[0181] Use P(θ) to represent the initial probability that the hypothesis θ has before the training data D. P(θ) is also called the prior probability of θ. The prior probability reflects the background knowledge about θ being the correct hypothesis. Without this prior knowledge, each candidate hypothesis can be simply assigned the same prior probability. Similarly, P(D) represents the probability distribution of the training data D, and P(D|θ) represents the probability of the training data D when the hypothesis θ holds. In machine learning, what needs to be concerned about is P(θ|D), that is, the probability that θ holds given the training data D. The result of Bayesian inference depends to a large extent on the prior probability. In addition, instead of completely accepting or rejecting the hypothesis, the likelihood of the hypothesis is only increased or decreased after observing more data.

[0182] It should be noted that the prior probability follows the prior distribution, and the posterior probability follows the posterior distribution. In this application, the method of moment estimation or the method of parameter solution can be used to obtain the prior distribution parameters corresponding to the prior distribution.

[0183] 103. Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution, where the target time period belongs to a time period within the historical time period;

[0184] In this embodiment, the advertisement placement device substitutes the calculated prior distribution parameters and the return conversion number information into the Bayesian estimation formula corresponding to the posterior distribution, and solves to obtain the target return ratio corresponding to the target advertisement in the target time period. Among them, the target time period belongs to a time period within the historical time period. Assuming that the historical time period is 24 hours and the target time period is the 5th hour, it is necessary to obtain the return conversion number corresponding to the 5th hour from the return conversion number information, and then substitute it into the Bayesian estimation formula to solve the target return ratio for the 5th hour.

[0185] 104. Determine the estimated conversion number of the target advertisement according to the target return ratio and the return conversion number information;

[0186] In this embodiment, after the advertisement placement device determines the target return ratio corresponding to the target advertisement, first obtain the return conversion number corresponding to the target time period from the return conversion number information, and then calculate the estimated conversion number of the target advertisement in the following manner:

[0187]

[0188] Among them, conv ht represents the estimated number of conversions corresponding to the target advertisement in the t-th time period, and rc ht represents the number of return conversions corresponding to the target advertisement in the t-th time period, and ratio ht represents the target return ratio corresponding to the target advertisement in the t-th time period.

[0189] 105. Determine the sorting result of the target advertisement among P advertisements according to the estimated number of conversions;

[0190] In this embodiment, since the target cost is a determined value, after the advertisement placement device determines the estimated number of conversions, the future cost achievement adjustment coefficient can be obtained, and the eCPM of the advertisement can be calculated based on the future cost achievement adjustment coefficient. Suppose it is necessary to sort 100 advertisements to be placed, then according to the eCPM corresponding to each advertisement to be placed, determine the sorting result of the target advertisement.

[0191] Specifically, for the convenience of introduction, please refer to Figure 8 , Figure 8 is a schematic diagram of an embodiment for adjusting the sorting of the target advertisement in the embodiment of the present application. As shown in Figure 8 figure (a) therein, assume that the current time is 17:00. At this time, the advertisements displayed on the "XXXX Flagship Store" include "Dot Shirt", "Bear T-shirt", "White-collar Trousers", "Kitten T-shirt", "Striped Shirt" and "All-black Belt". After one hour, the advertisement placement device re-sorts the advertisements in the "XXXX Flagship Store" based on the estimated number of conversions corresponding to this one hour. As shown in Figure 8 figure (b) therein, assume that the current time is 18:00. The advertisements displayed on the "XXXX Flagship Store" include "Kitten T-shirt", "Striped Shirt", "Beige Work Trousers", "Black and White T-shirt", "All-black Belt" and "Bear T-shirt".

[0192] 106. If it is determined that the target advertisement meets the advertisement placement conditions according to the sorting result, then place the target advertisement.

[0193] In this embodiment, the advertisement placement device needs to determine whether the sorting result of the target advertisement meets the advertisement placement conditions. Specifically, one determination condition is that, assuming the sorting result is less than or equal to the sorting threshold, it means that the target advertisement meets the advertisement placement conditions; assuming the sorting result is greater than the sorting threshold, it means that the target advertisement does not meet the advertisement placement conditions. For example, if the sorting result is 15 and the sorting threshold is 20, then the target advertisement meets the advertisement placement conditions. Another determination condition is that, assuming the sorting result is the first, it means that the target advertisement meets the advertisement placement conditions. There is also a determination condition that determines whether the target advertisement meets the advertisement placement conditions according to the selection of the advertiser. If the advertiser selects the target advertisement, it means that the target advertisement meets the advertisement placement conditions; if the advertiser does not select the target advertisement, it means that the target advertisement does not meet the advertisement placement conditions. It can be understood that in practical applications, other advertisement placement conditions can also be set according to the situation. This is only an illustration here and should not be construed as a limitation of the present application.

[0194] The present application provides a method for advertisement placement. First, obtain the return ratio information corresponding to P advertisements in the historical time period and the return conversion number information corresponding to the target advertisement in the historical time period. Then, determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information. Next, based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution. According to the target return ratio and the return conversion number information, determine the estimated conversion number of the target advertisement. Finally, determine the sorting result of the target advertisement among the P advertisements according to the estimated conversion number. If it is determined that the target advertisement meets the advertisement placement conditions according to the sorting result, then the target advertisement can be placed. Through the above method, the return ratio information statistically obtained for the same advertiser in the historical time period is used as the basis for constructing the prior probability. This return ratio information belongs to coarse-grained information, and the return conversion number information of the target advertisement is used as sample information, which belongs to fine-grained information. Based on Bayes' theorem, the posterior distribution can be derived according to the prior probability and the sample information, and then the target return ratio of the target advertisement can be determined based on the posterior distribution. Thus, introducing coarse-grained information as prior knowledge and fully considering the dependence relationship between coarse-grained information and fine-grained information, even if the fine-grained information is sparse, it is possible to calculate a target return ratio that is closer to the real situation, thereby improving the accuracy of the estimated conversion number, facilitating increasing the accuracy of advertisement sorting, and enhancing the advertisement placement effect.

[0195] Optionally, based on the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiment of the present application, obtaining the return ratio information corresponding to P advertisements in the historical time period and the return conversion number information corresponding to the target advertisement in the historical time period may include the following steps:

[0196] Obtain the return ratio information corresponding to P advertisements within a historical time period. The historical time period includes N time periods. The return ratio information includes N return ratios. The return ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1.

[0197] Obtain the return conversion number information corresponding to a target advertisement within a historical time period. The return conversion number information includes N return conversion numbers. The return conversion number has a corresponding relationship with the time period, and there are N return conversion numbers.

[0198] Among them, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement. A low-frequency advertisement is an advertisement for which the sum of N return conversion numbers is less than or equal to a return conversion number threshold, and a non-low-frequency advertisement is an advertisement for which the sum of N return conversion numbers is greater than the return conversion number threshold.

[0199] In this embodiment, a method for obtaining coarse-grained information and fine-grained information is introduced, which will be described separately below.

[0200] I. Coarse-grained information;

[0201] The coarse-grained information is specifically the return ratio information corresponding to P advertisements within a historical time period. The advertisement delivery device counts the return ratio information of P advertisements within a historical time period. The historical time period includes N time periods. For ease of understanding, please refer to Figure 9 , Figure 9 This is a schematic diagram of the return ratios corresponding to N time periods within the historical time period in the embodiment of the present application. As shown in the figure, assuming that the historical time period includes 24 time periods, the return ratio corresponding to each time period can be obtained. Please refer to Table 5, which is a schematic diagram of the return ratio information of P advertisements within a historical time period.

[0202] Table 5

[0203] Time period Return ratio Time period Return ratio 2 0.35 14 0.12 4 0.24 16 0.11 6 0.15 18 0.10 8 0.14 20 0.11 10 0.15 22 0.10 12 0.12 24 0.04

[0204] As can be seen from Table 5, the return ratio information includes N return ratios. Here, N is 12. In practical applications, N can also take other values, such as 6, 8, 24, or other values, etc. This is only a schematic diagram and should not be construed as a limitation to the present application.

[0205] II. Fine-grained information;

[0206] The fine-grained information is specifically the return conversion number information. The advertisement delivery device obtains the return conversion number information corresponding to the target advertisement reported by the advertiser within a historical time period. The historical time period includes N time periods. Assuming that the historical time period includes 24 time periods, the return conversion number corresponding to each time period can be obtained. Please refer to Table 6, which is a schematic diagram of the return conversion number information of the target advertisement within a historical time period.

[0207] Table 6

[0208] Time period Number of conversions from return Time period Number of conversions from return 2 2 14 1 4 1 16 2 6 0 18 0 8 1 20 1 10 0 22 0 12 1 24 0

[0209] As can be seen from Table 5, the reflux ratio information includes N reflux ratios, where N is 12. In practical applications, N can also take other values, such as 6, 8, 24, or other values. This is only for illustration and should not be construed as a limitation to this application.

[0210] Based on Table 6, it can be seen that the number of loop conversions of the target advertisement in each time period is relatively small. The sum of these reflux conversion numbers is 2 + 1 + 0 + 1 + 0 + 1 + 1 + 2 + 0 + 1 + 0 + 0 = 9. Assuming that the reflux conversion number threshold is 10, then the sum of these N reflux conversion numbers is less than the reflux conversion number threshold. Therefore, it can be considered that the target advertisement belongs to a low-frequency advertisement. Conversely, if the sum of the N reflux conversion numbers is greater than the reflux conversion number threshold, then the target advertisement belongs to a non-low-frequency advertisement.

[0211] Secondly, in the embodiments of the present application, a method for obtaining coarse-grained information and fine-grained information is provided. Through the above method, on the one hand, the coarse-grained information can be statistically obtained and the coarse-grained information is divided into reflux ratios corresponding to multiple time periods. On the other hand, the fine-grained information of low-frequency advertisements or non-low-frequency advertisements can be statistically obtained, and the coarse-grained information is also divided into reflux conversion numbers corresponding to multiple time periods, so as to facilitate the subsequent calculation of the reflux ratio corresponding to a certain time period, thereby improving the feasibility and operability of the solution.

[0212] Optionally, on the basis of the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiments of the present application, the prior distribution is the first beta distribution;

[0213] Determining the prior distribution parameters corresponding to the prior distribution according to the reflux ratio information may include the following steps:

[0214] Calculating the average value and variance according to the reflux ratio information;

[0215] Determining the first prior parameter in the prior distribution parameters according to the average value and variance;

[0216] Determining the second prior parameter in the prior distribution parameters according to the average value and variance, where the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

[0217] In this embodiment, a method for calculating the prior distribution parameters based on the first beta distribution is introduced. After obtaining the return ratio information corresponding to P advertisements in the historical time period, the prior distribution parameters of the first beta distribution can be obtained by using the method of moment estimation. Among them, the prior distribution parameters include the first prior parameter (α) and the second prior parameter (β). Suppose the calculated first prior parameter (α) is 30 and the calculated second prior parameter (β) is 70. For the convenience of understanding, please refer to Figure 10 , Figure 10 is a schematic diagram of a posterior probability density based on the first beta distribution in the embodiment of the present application. As Figure 10 shown, the first beta distribution indicated by S1 in the figure is the first beta distribution of the prior probability, denoted as beta(30, 70). Based on this, the return conversion number information of multiple advertisements can be obtained. Please refer to Table 7. Table 7 is a schematic diagram of the return conversion number information of Advertisement A, Advertisement B, and Advertisement C on the first day.

[0218] Table 7

[0219] Number of conversions from first-day return Total number of conversions First-day return ratio Advertisement A 1 2 50％ Advertisement B 2 2 100％ Advertisement C 240 300 80％ Advertiser * Product identifier 300 1000 30％

[0220] Combined with Figure 10 and Table 7, it can be seen that the total conversion number of Advertisement A is 2, and the return conversion number on the first day is 1. Then the conversion number that did not return on the first day is 1, which means that the return was successful 1 time and the return failed 1 time. Therefore, the first beta distribution of Advertisement A is denoted as beta(31, 71), where 31 is α + 1 and 71 is β + 1. Similarly, the total conversion number of Advertisement B is 2, and the return conversion number on the first day is 2. Then the conversion number that did not return on the first day is 0, which means that the return was successful 2 times and the return failed 0 times. Therefore, the first beta distribution of Advertisement B is denoted as beta(32, 70), where 32 is α + 2 and 70 is β + 0. For Advertisement C, the total conversion number is 300, and the return conversion number on the first day is 240. Then the conversion number that did not return on the first day is 60, which means that the return was successful 240 times and the return failed 60 times. Therefore, the first beta distribution of Advertisement C is denoted as beta(270, 130), where 270 is α + 240 and 130 is β + 60.

[0221] It can be seen that the return conversion number and the total conversion number on the first day of Advertisement A and Advertisement B are both relatively sparse, that is, both Advertisement A and Advertisement B are low-frequency advertisements. Among them, the first beta distribution curve S2 shown by Advertisement A is relatively close to the first beta distribution curve S1 of the prior probability, and the first beta distribution curve S3 shown by Advertisement B is also relatively close to the first beta distribution curve S1 of the prior probability, and the posterior distribution approaches the prior distribution, indicating that for low-frequency advertisements, they rely more on coarse-grained information. For Advertisement C, the data is relatively sufficient, and the posterior approaches the likelihood estimation.

[0222] Taking the method of moment estimation as an example, this section introduces how to calculate the first prior parameter (α) and the second prior parameter (β) corresponding to the first beta distribution. Here, moment estimation is an approximate estimation method for parameters. Its basic idea is to use sample moments to estimate population moments. According to the law of large numbers, if the unknown parameter is related to a certain (some) moment of the population, then an estimate of the unknown parameter can be constructed. Moments include the first moment, the second moment, and higher-order moments, etc. The commonly used ones are the first and second moments. The first moment is also called the static moment, which is the integral (for continuous functions) or summation (for discrete functions) of the product of the function and the independent variable xf(x). In mechanics, it is used to represent the resultant moment of the distributed force f(x) to a certain point, and geometrically it can be used to calculate the centroid, which is called the mathematical expectation (or mean) in statistics. There is also the second central moment (or variance) in statistics. The specific calculation steps are as follows:

[0223] First, according to the given probability density function, calculate the origin moments of the population. If there is only one parameter, only the first origin moment needs to be calculated. If there are two parameters, then the first and second moments need to be calculated. Assuming there are two parameters, the mathematical expectation needs to be calculated first:

[0224]

[0225] Then calculate the variance:

[0226]

[0227] where x represents the return ratio of each of the P advertisements per hour within N hours.

[0228] Based on this, then according to the given training data (i.e., the return ratio of each of the P advertisements per hour within N hours), calculate the origin moments of the training data. Make the origin moments of the population equal to the origin moments of the training data, and solve for the parameters. The resulting value is the moment estimate value of the parameter. At this time, in the first beta distribution, it can be calculated that:

[0229] mean = E(x) = α / (α + β);

[0230] var = D(x) = αβ / (α + β) 2 (α + β + 1);

[0231] So the solution is:

[0232] α = [mean × (1 - mean) / var - 1] × mean;

[0233] β = [mean × (1 - mean) / var - 1] × (1 - mean);

[0234] Wherein, α represents the first prior parameter corresponding to the first beta distribution, and β represents the second prior parameter corresponding to the first beta distribution.

[0235] Again, in the embodiments of the present application, a method for calculating prior distribution parameters based on the beta distribution is provided. Through the above method, the prior distribution is defined as the beta distribution. The beta distribution can reasonably estimate the return ratio of advertisements, and for both low-frequency advertisements and non-low-frequency advertisements, the target return ratio estimated based on the beta distribution has higher accuracy. Especially for low-frequency advertisements, the accuracy of the target return ratio has a more obvious improvement.

[0236] Optionally, based on the above Figure 8 On the basis of the corresponding embodiments, in another optional embodiment of the advertisement placement method provided in the embodiments of the present application, based on the prior distribution parameters and the return conversion number information, the target return ratio corresponding to the target advertisement in the target time period is determined through the posterior distribution, which may include the following steps:

[0237] Determine the binomial distribution according to the return conversion number information;

[0238] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution;

[0239] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0240] Wherein, determining the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution may include the following steps:

[0241] Determine the second beta distribution according to the binomial distribution and the first beta distribution, wherein the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution;

[0242] Determine the Bayesian estimation formula according to the second beta distribution.

[0243] In this embodiment, a method for deriving the Bayesian estimation formula using the first beta distribution and the binomial distribution (Binomial distribution) is introduced. Based on the above embodiments, it can be known that if the prior distribution is the first beta distribution, the prior distribution parameters of the first beta distribution are obtained by the method of moment estimation as (α, β), and the return conversion of the target advertisement satisfies the binomial distribution, that is Wherein, the return conversion number information of the target advertisement is And the return ratio information of the target advertisement is Based on this, the following relationship is obtained:

[0244]

[0245] Among them, B(α,β) represents the first beta distribution and belongs to the prior distribution. represents the binomial distribution and belongs to the sample information. B(α + c1, β + c2) represents the second beta distribution and belongs to the posterior distribution. The symbol represents being derived from the previous formula.

[0246] Combined with the Bayesian formula P(θ|D) ∝ P(D|θ) × P(θ), the Bayesian estimation formula can be obtained:

[0247]

[0248] Among them, r1 represents the return ratio of the target advertisement in the first time period, and r2 represents the return ratio of the target advertisement in the second time period. c1 represents the return conversion number of the target advertisement in the first time period, c2 represents the return conversion number of the target advertisement in the second time period, α represents the first prior parameter, and β represents the second prior parameter.

[0249] It should be noted that the beta distribution (i.e., the first beta distribution in this application) is the conjugate prior distribution of the binomial distribution. Usually, it can be assumed that the prior probability conforms to a certain law or distribution, and then according to the additional information, the calculation formula or distribution of the posterior probability can also be obtained. If the prior probability and the posterior probability conform to the same distribution, then this distribution is called the conjugate distribution. The advantage of the conjugate distribution is that it can clearly show the influence of the newly added information on the distribution parameters, that is, the change law of the probability distribution.

[0250] Furthermore, in the embodiments of this application, a method for deriving the Bayesian estimation formula using the first beta distribution and the binomial distribution is provided. Through the above method, since the beta distribution is the conjugate distribution of the binomial distribution, that is to say, if the return conversion number information conforms to the binomial distribution, then both the prior distribution and the posterior distribution can maintain the form of the beta distribution. Thus, the Bayesian estimation formula can be derived from the posterior distribution, so as to achieve the purpose of obtaining the target return ratio based on theoretical knowledge, which is beneficial to the feasibility and operability of the solution.

[0251] Optionally, on the basis of the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiments of this application, based on the prior distribution parameters and the return conversion number information, to determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula, the following steps can be included:

[0252] Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to a return conversion number in the return conversion number information;

[0253] Based on the first prior parameter, the second prior parameter, and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0254] In this embodiment, a specific method for calculating the target return ratio based on the Bayesian estimation formula is introduced. As can be seen from the above embodiment, the Bayesian estimation formula is determined based on the first beta distribution and the binomial distribution, and the Bayesian estimation formula can be expressed as:

[0255]

[0256] Among them, α represents the first prior parameter, β represents the second prior parameter. Here, the first prior parameter α and the second prior parameter β can be calculated through moment estimation, which will not be elaborated here. In addition, it is also necessary to obtain the target return conversion number corresponding to the target advertisement in the target time period. Assuming that the target time period is the 5th hour (i.e., i = 5), then extract the return conversion number corresponding to the 5th hour according to the return conversion number information, that is, obtain the target return conversion number c5, and then calculate the target return ratio r5.

[0257] It should be noted that the method for calculating the return ratio corresponding to other time periods is similar to the above method, so it will not be listed one by one here.

[0258] Furthermore, in the embodiment of the present application, a specific method for calculating the target return ratio based on the Bayesian estimation formula is provided. Through the above method, substitute the calculated first prior parameter and second prior parameter into the Bayesian estimation formula, and substitute a return conversion number in the return conversion number information into the Bayesian estimation formula, so as to calculate the target return ratio, thereby improving the feasibility and operability of the solution.

[0259] Optionally, on the basis of the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiment of the present application, the prior distribution is the first Dirichlet distribution;

[0260] Determining the prior distribution parameters corresponding to the target advertisement according to the return ratio information may include the following steps:

[0261] Calculate the average value and variance according to the return ratio information;

[0262] Determine N prior parameters in the prior distribution parameters according to the mean value and variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

[0263] In this embodiment, a method for calculating prior distribution parameters based on the Dirichlet distribution is introduced. After obtaining the return ratio information corresponding to P advertisements in the historical time period, the prior distribution parameters of the first Dirichlet distribution can be obtained by using the method of moment estimation first, where the prior distribution parameters include N prior parameters. The specific calculation steps are as follows:

[0264] First, according to the given probability density function, calculate the origin moment of the population. If there is only one parameter, only the first-order origin moment needs to be calculated. If there are two parameters, the first-order and second-order moments need to be calculated. Assuming there are two parameters, the mathematical expectation needs to be calculated first:

[0265]

[0266] Then calculate the variance:

[0267]

[0268] Where x represents the return ratio of each of the P advertisements per hour within N hours.

[0269] Based on this, then according to the given training data (i.e., the return ratio of each of the P advertisements per hour within N hours), calculate the origin moment of the training data. Make the origin moment of the population equal to the origin moment of the training data, and solve for the parameters. The obtained result is the moment estimation value of the parameters. At this time, in the first Dirichlet distribution, it can be calculated that:

[0270]

[0271] Where E(x i ) represents the expectation of the return ratio in the i-th time period, D(x i ) represents the variance of the return ratio in the i-th time period, c i represents the number of return conversions in the i-th time period, c n represents the number of return conversions in the n-th time period. Then the prior distribution parameters are solved

[0272] Again, in the embodiments of the present application, a method for calculating the prior distribution parameters based on the Dirichlet distribution is provided. Through the above method, the prior distribution is defined as the Dirichlet distribution, and the Dirichlet distribution can reasonably estimate the return ratio of the advertisement. Moreover, for low-frequency advertisements and non-low-frequency advertisements, the target return ratios estimated based on the Dirichlet distribution are all more accurate. Especially for low-frequency advertisements, the accuracy of the target return ratio is significantly improved.

[0273] Optionally, based on the corresponding embodiments above, in another optional embodiment of the advertisement placement method provided in the embodiments of the present application, based on the prior distribution parameters and the return conversion number information, the target return ratio corresponding to the target advertisement in the target time period is determined through the posterior distribution, which may include the following steps: Figure 8 Based on the prior distribution parameters and the return conversion number information, the target return ratio corresponding to the target advertisement in the target time period is determined through the posterior distribution, which may include the following steps:

[0274] Determine the multinomial distribution according to the return conversion number information;

[0275] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution;

[0276] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0277] Among them, determining the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution may include the following steps:

[0278] Determine the second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution;

[0279] Determine the Bayesian estimation formula according to the second Dirichlet distribution.

[0280] In this embodiment, a method for deriving the Bayesian estimation formula using the first Dirichlet distribution and the multinomial distribution is introduced. Based on the above embodiments, it can be known that if the prior distribution is the first Dirichlet distribution, the prior distribution parameters of the first Dirichlet distribution are obtained by the method of moment estimation as And the return conversion of the target advertisement satisfies the multinomial distribution, that is Among them, the return conversion number information of the target advertisement is And the return ratio information of the target advertisement is N represents N time periods. Based on this, the following relationship is obtained:

[0281]

[0282] Among them, represents the first Dirichlet distribution and belongs to the prior distribution, represents the multinomial distribution and belongs to the sample information, represents the second Dirichlet distribution and belongs to the posterior distribution. The symbol means derived from the previous formula.

[0283] Combined with the Bayesian formula P(θ|D) ∝ P(D|θ) × P(θ), the Bayesian estimation formula can be obtained:

[0284]

[0285] Among them, r i represents the return ratio of the target advertisement in the i-th hour, that is, the target return ratio corresponding to the target advertisement in the target time period is obtained. The i-th hour represents the target time period, and c i represents the return conversion number in the i-th hour, and a i represents the prior distribution parameter the i-th parameter in.

[0286] It should be noted that the Dirichlet distribution (i.e., the first Dirichlet distribution in this application) is the conjugate prior distribution of the multinomial distribution. Usually, it can be assumed that the prior probability conforms to a certain law or distribution, and then according to the additional information, the calculation formula or distribution of the posterior probability can also be obtained. If the prior probability and the posterior probability conform to the same distribution, then this distribution is called the conjugate distribution. The advantage of the conjugate distribution is that it can clearly show the influence of the newly added information on the distribution parameters, that is, the change law of the probability distribution.

[0287] Furthermore, in the embodiments of this application, a method for deriving the Bayesian estimation formula using the first Dirichlet distribution and the multinomial distribution is provided. Through the above method, since the Dirichlet distribution is the conjugate distribution of the multinomial distribution, that is to say, if the return conversion number information conforms to the multinomial distribution, then both the prior distribution and the posterior distribution can maintain the form of the Dirichlet distribution. Thus, the Bayesian estimation formula can be derived from the posterior distribution, so as to achieve the purpose of obtaining the target return ratio based on theoretical knowledge, which is beneficial to the feasibility and operability of the solution.

[0288] Optionally, based on the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiments of this application, based on the prior distribution parameter and the return conversion number information, determining the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula may include the following steps:

[0289] Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to one of the return conversion numbers in the return conversion number information;

[0290] Based on N prior parameters and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0291] In this embodiment, a specific method for calculating the target return ratio based on the Bayesian estimation formula is introduced. As can be seen from the above embodiment, the Bayesian estimation formula is determined based on the first Dirichlet distribution and the multinomial distribution, and the Bayesian estimation formula can be expressed as:

[0292]

[0293] where α represents the i-th parameter in the prior distribution parameter Among them, the N prior parameters can be calculated through moment estimation, which will not be elaborated here. In addition, it is also necessary to obtain the target return conversion number corresponding to the target advertisement in the target time period. Assuming that the target time period is the 5th hour (i.e., i = 5), then extract the return conversion number corresponding to the 5th hour according to the return conversion number information, that is, obtain the target return conversion number c5, and then calculate the target return ratio r5.

[0294] It should be noted that the method for calculating the return ratio corresponding to other time periods is similar to the above method, so it will not be listed one by one here.

[0295] Furthermore, in the embodiment of the present application, a specific method for calculating the target return ratio based on the Bayesian estimation formula is provided. Through the above method, substitute the calculated N prior parameters into the Bayesian estimation formula, and substitute one of the return conversion numbers in the return conversion number information into the Bayesian estimation formula, so as to calculate the target return ratio, thereby improving the feasibility and operability of the solution.

[0296] Optionally, on the basis of the above Figure 8 corresponding embodiment, in another optional embodiment of the advertisement placement method provided in the embodiment of the present application, determining the sorting result of the target advertisement among P advertisements according to the estimated conversion number may include the following steps:

[0297] Determine the future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number;

[0298] Determine the revenue per thousand impressions corresponding to the target advertisement according to the future cost achievement adjustment coefficient corresponding to the target advertisement;

[0299] Obtain the revenue per thousand impressions corresponding to each advertisement to be placed among the P advertisements, where the revenue per thousand impressions corresponding to each advertisement to be placed is determined according to the future cost achievement adjustment coefficient corresponding to each advertisement to be placed;

[0300] Sort the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each advertisement to be placed to obtain the sorting result of the target advertisement among the P advertisements.

[0301] In this embodiment, a method for automatically sorting P advertisements according to the estimated conversion number is introduced. Usually, one or more target time periods can be set first. For example, taking the historical time period as 24 hours as an example, assuming the target time period is the 5th hour, then according to the target return ratio at the 5th hour and the return conversion number at the 5th hour, the estimated conversion number corresponding to the 5th hour is calculated. Also assume that the multiple target time periods are the 5th hour and the 6th hour respectively. Then, according to the target return ratio at the 5th hour and the return conversion number at the 5th hour, the estimated conversion number corresponding to the 5th hour is calculated, and according to the target return ratio at the 6th hour and the return conversion number at the 6th hour, the estimated conversion number corresponding to the 6th hour is calculated. Then, taking the average of the return conversion number at the 5th hour and the estimated conversion number corresponding to the 6th hour, the estimated conversion number of the target advertisement required can be obtained. It can be understood that a similar method can also be used to estimate other advertisements of the same advertisement publisher, and the estimated conversion numbers corresponding to these advertisements are obtained respectively. Among them, the advertisement publisher can specifically be some media platforms or enterprises that place advertisements.

[0302] Specifically, the advertisement publisher can also set a target cost in advance. Since the target cost is ultimately only related to the estimated conversion number within the historical time period and the future cost achievement adjustment coefficient, after determining the target cost and the estimated conversion number within the historical time period, the future cost achievement adjustment coefficient can be obtained, and thus the revenue per thousand impressions of the P advertisements can be calculated. Finally, the advertisement placement device can generate a sorting result according to the revenue per thousand impressions of these advertisements.

[0303] In practical applications, the advertisement placement device pushes the sorting result of the P advertisements to the client used by the advertisement publisher, providing the function of sorting and displaying for the advertisement publisher, which is convenient for the advertisement publisher to plan the placement plan in a timely manner. For the convenience of introduction, please refer to Figure 11 , Figure 11 is a schematic diagram of an interface for pushing the sorting result of the target advertisement in the embodiment of the present application. As shown in the figure, assume that the advertisement publisher XXX logs in to the advertisement placement interface using an account and password, and selects to view the advertisement sorting result through the advertisement placement interface. The sorting result can be arranged in ascending order or in descending order. Figure 11It is introduced by taking the descending order of the revenue per thousand impressions as an example, however, this should not be construed as a limitation to this application. The advertising publisher can view the revenue per thousand impressions of different advertisements, and can also view the identifier of each advertisement and the corresponding product identifier, and can also view the product content corresponding to the product identifier. Through Figure 11 It can be seen that the advertisement with the highest current revenue per thousand impressions is the advertisement with the advertisement identifier "025", and the product content is "marshmallow", while the advertisement with the lowest current revenue per thousand impressions is the advertisement corresponding to the advertisement identifier "039", and the product content is "skittles".

[0304] Secondly, in the embodiment of this application, a method for automatically sorting P advertisements according to the estimated conversion number is provided. Through the above method, a more accurate future cost achievement adjustment coefficient can be estimated using a more accurate estimated conversion number, thereby determining the eCPM of the target advertisement. A similar method can be used to obtain the eCPM of other advertisements placed by the same advertising publisher, and the eCPM of these P advertisements are automatically sorted in a certain order to obtain the sorting result of the P advertisements. This can not only provide a more accurate sorting result for the advertising publisher, but also facilitate the advertising publisher to plan subsequent advertising placement plans.

[0305] Optionally, based on the above Figure 8 corresponding embodiment, in another optional embodiment of the advertising placement method provided in the embodiment of this application, the following steps may further be included:

[0306] Push the sorting result of the target advertisement among the P advertisements to the client;

[0307] Receive an advertisement placement selection instruction sent by the client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, where Q is an integer greater than or equal to 1 and less than or equal to P;

[0308] If the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement, it is determined that the target advertisement meets the advertisement placement condition.

[0309] In this embodiment, a method for setting the advertisement placement situation is introduced. Based on the above embodiment, it can be known that after the advertising placement device calculates the revenue per thousand impressions corresponding to each of the P advertisements, it can push the sorting result of the P advertisements to the client used by the advertising publisher, that is, provide the function of sorting and displaying for the advertising publisher, which is convenient for the advertising publisher to plan the placement plan on the spot. For the convenience of introduction, please refer to Figure 12 , Figure 12This is a schematic diagram of an interface for triggering an advertisement placement selection instruction in an embodiment of the present application. As shown in the figure, it is assumed that the advertisement publisher XXX logs in to the advertisement placement interface using an account and password, and selects to view the advertisement sorting results through the advertisement placement interface. The sorting results can be arranged in ascending order or in descending order. Figure 12 The following takes the descending order of the revenue per thousand impressions as an example for introduction, but this should not be construed as a limitation to the present application. The advertisement publisher can view the revenue per thousand impressions of different advertisements, can also view the identifier of each advertisement and the corresponding product identifier, and can also view the product content corresponding to the product identifier. According to the revenue per thousand impressions corresponding to each advertisement, the advertisement publisher can check whether to continue placing the advertisement. It is assumed that the advertisement publisher selects to place the advertisement with the advertisement identifier "025" and the advertisement with the advertisement identifier "007", then an advertisement placement selection instruction is triggered. At this time, the advertisement placement selection instruction carries 2 advertisement identifiers (i.e., when Q is equal to 2), which are the advertisement identifier "025" and the advertisement identifier "007" respectively. It is assumed that the target advertisement is the advertisement with the advertisement identifier "025" or the advertisement with the advertisement identifier "007", that is, the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement. Then the target advertisement meets the advertisement placement conditions. If the Q advertisement identifiers do not include the advertisement identifier corresponding to the target advertisement, then the target advertisement does not meet the advertisement placement conditions.

[0310] Secondly, in an embodiment of the present application, a method for setting the advertisement placement situation is provided. Through the above method, the advertisement publisher can also select whether to place the target advertisement according to actual needs. Compared with directly determining whether to place the target advertisement according to the revenue per thousand impressions, adding an active selection placement scheme can increase the flexibility and feasibility of the scheme, thereby improving the practicality and operability of advertisement placement.

[0311] The advertisement placement device in the present application will be described in detail below. Please refer to Figure 13 , Figure 13 This is a schematic diagram of an embodiment of the advertisement placement device in an embodiment of the present application. The advertisement placement device 20 may include:

[0312] An acquisition module 201, configured to acquire the return ratio information corresponding to P advertisements in a historical time period, and the return conversion number information corresponding to the target advertisement in the historical time period, where the P advertisements at least include the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1;

[0313] A determination module 202, configured to determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information;

[0314] The determination module 202 is further configured to determine, based on the prior distribution parameters and the return conversion number information, a target return ratio corresponding to the target advertisement in a target time period through a posterior distribution, where the target time period belongs to a time period within the historical time period;

[0315] The determination module 202 is further configured to determine the estimated conversion number of the target advertisement according to the target return ratio and the return conversion number information;

[0316] The determination module 202 is further configured to determine the sorting result of the target advertisement among the P advertisements according to the estimated conversion number;

[0317] The placement module 203 is configured to place the target advertisement if it is determined according to the sorting result that the target advertisement meets the advertisement placement condition.

[0318] The present application provides an advertisement placement device. The advertisement placement device obtains the return ratio information corresponding to P advertisements in the historical time period and the return conversion number information corresponding to the target advertisement in the historical time period, then determines the prior distribution parameters corresponding to the prior distribution according to the return ratio information, and then determines the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution based on the prior distribution parameters and the return conversion number information. The estimated conversion number of the target advertisement is determined according to the target return ratio and the return conversion number information, and finally the sorting result of the target advertisement among the P advertisements is determined according to the estimated conversion number. If it is determined according to the sorting result that the target advertisement meets the advertisement placement condition, then the target advertisement can be placed. By using the above device, the return ratio information statistically obtained by the same advertiser in the historical time period is used as the basis for constructing the prior probability. The return ratio information belongs to coarse-grained information, and the return conversion number information of the target advertisement is used as sample information, which belongs to fine-grained information. Based on Bayes' theorem, the posterior distribution can be deduced according to the prior probability and the sample information, and then the target return ratio of the target advertisement can be determined based on the posterior distribution. Thus, the coarse-grained information is introduced as prior knowledge, and the dependence relationship between the coarse-grained information and the fine-grained information is fully considered. Even if the fine-grained information is sparse, the target return ratio closer to the actual situation can be calculated, thereby improving the accuracy of the estimated conversion number, being beneficial to increasing the accuracy of advertisement sorting, and improving the advertisement placement effect.

[0319] Optionally, based on the corresponding embodiment above, Figure 13 In another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0320] The acquisition module 201 is specifically configured to acquire the return ratio information corresponding to P advertisements in the historical time period, where the historical time period includes N time periods, the return ratio information includes N return ratios, the return ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1;

[0321] Obtain the return conversion number information corresponding to the target advertisement in the historical time period, where the return conversion number information includes N return conversion numbers, the return conversion numbers have a corresponding relationship with the time period, and N return conversion numbers;

[0322] Among them, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement. A low-frequency advertisement is an advertisement whose sum of N return conversion numbers is less than or equal to the return conversion number threshold, and a non-low-frequency advertisement is an advertisement whose sum of N return conversion numbers is greater than the return conversion number threshold.

[0323] Optionally, based on the above Figure 13 In another embodiment of the advertisement delivery device 20 provided by the embodiment of the present application on the basis of the corresponding embodiment, the prior distribution is the first beta distribution;

[0324] The determination module 202 is specifically configured to calculate the average value and variance according to the return ratio information;

[0325] Determine the first prior parameter in the prior distribution parameters according to the average value and variance;

[0326] Determine the second prior parameter in the prior distribution parameters according to the average value and variance, where the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

[0327] Optionally, based on the above Figure 13 In another embodiment of the advertisement delivery device 20 provided by the embodiment of the present application on the basis of the corresponding embodiment,

[0328] The determination module 202 is specifically configured to determine the binomial distribution according to the return conversion number information;

[0329] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution;

[0330] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0331] Optionally, based on the above Figure 13 In another embodiment of the advertisement delivery device 20 provided by the embodiment of the present application on the basis of the corresponding embodiment,

[0332] The determination module 202 is specifically configured to determine the second beta distribution according to the binomial distribution and the first beta distribution, where the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution;

[0333] Determine the Bayesian estimation formula according to the second beta distribution.

[0334] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0335] The determining module 202 is specifically configured to obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to a return conversion number in the return conversion number information;

[0336] Based on the first prior parameter, the second prior parameter, and the target return conversion number, the target return ratio corresponding to the target advertisement in the target time period is calculated through the Bayesian estimation formula.

[0337] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application, the prior distribution is the first Dirichlet distribution;

[0338] The determining module 202 is specifically configured to calculate the average value and the variance according to the return ratio information;

[0339] N prior parameters in the prior distribution parameters are determined according to the average value and the variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

[0340] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0341] The determining module 202 is specifically configured to determine the multinomial distribution according to the return conversion number information;

[0342] According to the multinomial distribution and the first Dirichlet distribution, the Bayesian estimation formula corresponding to the posterior distribution is determined;

[0343] Based on the prior distribution parameters and the return conversion number information, the target return ratio corresponding to the target advertisement in the target time period is determined through the Bayesian estimation formula.

[0344] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0345] The determining module 202 is specifically configured to determine the second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution;

[0346] The Bayesian estimation formula is determined according to the second Dirichlet distribution.

[0347] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0348] The determining module 202 is specifically configured to obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to one return conversion number in the return conversion number information;

[0349] Based on N prior parameters and the target return conversion number, the target return ratio corresponding to the target advertisement in the target time period is calculated through the Bayesian estimation formula.

[0350] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application,

[0351] The determining module 202 is specifically configured to determine the future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number;

[0352] According to the future cost achievement adjustment coefficient corresponding to the target advertisement, determine the revenue per thousand impressions corresponding to the target advertisement;

[0353] Obtain the revenue per thousand impressions corresponding to each to-be-placed advertisement among P advertisements, where the revenue per thousand impressions corresponding to each to-be-placed advertisement is determined according to the future cost achievement adjustment coefficient corresponding to each to-be-placed advertisement;

[0354] Sort the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each to-be-placed advertisement to obtain the sorting result of the target advertisement among the P advertisements.

[0355] Optionally, based on the above Figure 13 corresponding embodiment, in another embodiment of the advertisement placement device 20 provided by the embodiment of the present application, the advertisement placement device 20 further includes a push module 204 and a receiving module 205;

[0356] The push module 204 is configured to push the sorting result of the target advertisement among the P advertisements to the client;

[0357] The receiving module 205 is configured to receive an advertisement placement selection instruction sent by the client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, where Q is an integer greater than or equal to 1 and less than or equal to P;

[0358] The determining module 202 is further configured to determine that the target advertisement meets the advertisement placement condition if the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement.

[0359] The computer device provided in this application may specifically be a server or a terminal device. This application takes the computer device as a server as an example for introduction. Please refer to Figure 14 , Figure 14 FIG. is a schematic structural diagram of a computer device provided by an embodiment of this application. The computer device 300 may vary greatly due to different configurations or performances, and may include one or more central processing units (CPUs) 322 (for example, one or more processors) and a memory 332, and one or more storage media 330 (for example, one or more mass storage devices) for storing application programs 342 or data 344. Among them, the memory 332 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the computer device. Further, the central processing unit 322 may be set to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the computer device 300.

[0360] The computer device 300 may further include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input / output interfaces 358, and / or one or more operating systems 341, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM , FreeBSD TM and so on.

[0361] In the embodiment of this application, the CPU 322 included in the computer device further has the following functions:

[0362] Obtain the return ratio information corresponding to P advertisements in the historical time period, and the return conversion number information corresponding to the target advertisement in the historical time period, where the P advertisements at least include the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1;

[0363] Determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information;

[0364] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution, where the target time period belongs to a time period within the historical time period;

[0365] Determine the estimated conversion number of the target advertisement according to the target return flow ratio and the return flow conversion number information;

[0366] Determine the sorting result of the target advertisement among the P advertisements according to the estimated conversion number;

[0367] If it is determined that the target advertisement meets the advertisement placement condition according to the sorting result, then place the target advertisement.

[0368] Optionally, the CPU 322 is specifically used to implement the following steps:

[0369] Obtain the return flow ratio information corresponding to the P advertisements in the historical time period. Among them, the historical time period includes N time periods, the return flow ratio information includes N return flow ratios, the return flow ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1;

[0370] Obtain the return flow conversion number information corresponding to the target advertisement in the historical time period. Among them, the return flow conversion number information includes N return flow conversion numbers, the return flow conversion number has a corresponding relationship with the time period, and N return flow conversion numbers;

[0371] Among them, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement. A low-frequency advertisement is an advertisement whose sum of N return flow conversion numbers is less than or equal to the return flow conversion number threshold, and a non-low-frequency advertisement is an advertisement whose sum of N return flow conversion numbers is greater than the return flow conversion number threshold.

[0372] Optionally, the CPU 322 is specifically used to implement the following steps:

[0373] Calculate the average value and variance according to the return flow ratio information;

[0374] Determine the first prior parameter in the prior distribution parameters according to the average value and variance;

[0375] Determine the second prior parameter in the prior distribution parameters according to the average value and variance. Among them, the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

[0376] Optionally, the CPU 322 is specifically used to implement the following steps:

[0377] Determine the binomial distribution according to the return flow conversion number information;

[0378] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution;

[0379] Based on the prior distribution parameters and the return flow conversion number information, determine the target return flow ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0380] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0381] Determine a second beta distribution according to the binomial distribution and the first beta distribution, where the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution;

[0382] Determine the Bayesian estimation formula according to the second beta distribution.

[0383] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0384] Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number is one of the return conversion number information;

[0385] Based on the first prior parameter, the second prior parameter, and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0386] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0387] Determine the prior distribution parameters corresponding to the target advertisement according to the return ratio information, including:

[0388] Calculate the average value and variance according to the return ratio information;

[0389] Determine N prior parameters in the prior distribution parameters according to the average value and variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

[0390] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0391] Determine the multinomial distribution according to the return conversion number information;

[0392] Determine the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution;

[0393] Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0394] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0395] Determine a second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution;

[0396] Determine the Bayesian estimation formula according to the second Dirichlet distribution.

[0397] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0398] Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to one of the return conversion number information;

[0399] Based on N prior parameters and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

[0400] Optionally, the CPU 322 is specifically configured to implement the following steps:

[0401] Determine the future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number;

[0402] Determine the revenue per thousand impressions corresponding to the target advertisement according to the future cost achievement adjustment coefficient corresponding to the target advertisement;

[0403] Obtain the revenue per thousand impressions corresponding to each to-be-delivered advertisement among P advertisements, where the revenue per thousand impressions corresponding to each to-be-delivered advertisement is determined according to the future cost achievement adjustment coefficient corresponding to each to-be-delivered advertisement;

[0404] Sort the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each to-be-delivered advertisement to obtain the sorting result of the target advertisement among the P advertisements.

[0405] Optionally, the CPU 322 is further configured to implement the following steps:

[0406] Push the sorting result of the target advertisement among the P advertisements to the client;

[0407] Receive an advertisement placement selection instruction sent by the client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, where Q is an integer greater than or equal to 1 and less than or equal to P;

[0408] If the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement, determine that the target advertisement meets the advertisement placement condition.

[0409] In an embodiment of the present application, a computer-readable storage medium is further provided, in which a computer program is stored, and when it runs on a computer, it causes the computer to execute the method described in the previous embodiment.

[0410] In an embodiment of the present application, a computer program product including a program is further provided, and when it runs on a computer, it causes the computer to execute the method described in the previous embodiment.

[0411] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0412] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0413] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0414] In addition, the functional units in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0415] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0416] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for advertising placement, characterized in that, Including: Obtaining the return ratio information corresponding to P advertisements within a historical time period, and the return conversion number information corresponding to a target advertisement within the historical time period, where the P advertisements at least include the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1; the return ratio information represents the ratio of the return conversion number to the total conversion number; the return conversion number information represents the converted number that has been returned; Determining the prior distribution parameters corresponding to the prior distribution according to the return ratio information; Based on the prior distribution parameters and the return conversion number information, determining the target return ratio corresponding to the target advertisement in a target time period through a posterior distribution, where the target time period belongs to a time period within the historical time period; Determining the estimated conversion number of the target advertisement according to the target return ratio and the return conversion number information; Determining the sorting result of the target advertisement among the P advertisements according to the estimated conversion number; If it is determined according to the sorting result that the target advertisement meets the advertisement placement condition, then placing the target advertisement.

2. The method according to claim 1, characterized in that The obtaining the return ratio information corresponding to P advertisements within a historical time period, and the return conversion number information corresponding to a target advertisement within the historical time period includes: Obtaining the return ratio information corresponding to the P advertisements within the historical time period, where the historical time period includes N time periods, the return ratio information includes N return ratios, the return ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1; Obtaining the return conversion number information corresponding to the target advertisement within the historical time period, where the return conversion number information includes N return conversion numbers, the return conversion number has a corresponding relationship with the time period, and the N return conversion numbers; Wherein, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement, the low-frequency advertisement is an advertisement whose sum of the N return conversion numbers is less than or equal to a return conversion number threshold, and the non-low-frequency advertisement is an advertisement whose sum of the N return conversion numbers is greater than the return conversion number threshold.

3. The method according to claim 1 or 2, characterized in that, The prior distribution is a first beta distribution; The determining the prior distribution parameters corresponding to the prior distribution according to the return ratio information includes: Calculating an average value and a variance according to the return ratio information; Determining a first prior parameter in the prior distribution parameters according to the average value and the variance; Determining a second prior parameter in the prior distribution parameters according to the average value and the variance, where the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

4. The method according to claim 3, wherein The determining the target return ratio corresponding to the target advertisement in a target time period through a posterior distribution based on the prior distribution parameters and the return conversion number information includes: Determining a binomial distribution according to the return conversion number information; Determining a Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution; Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

5. The method according to claim 4, wherein The determining of the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution includes: Determine a second beta distribution according to the binomial distribution and the first beta distribution, where the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution; Determine the Bayesian estimation formula according to the second beta distribution.

6. The method according to claim 4, wherein The determining of the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula based on the prior distribution parameters and the return conversion number information includes: Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number is one of the return conversion numbers in the return conversion number information; Calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula based on the first prior parameter, the second prior parameter, and the target return conversion number.

7. The method according to claim 1 or 2, characterized in that, The prior distribution is the first Dirichlet distribution; The determining of the prior distribution parameters corresponding to the target advertisement according to the return ratio information includes: Calculate the mean value and variance according to the return ratio information; Determine N prior parameters in the prior distribution parameters according to the mean value and the variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

8. The method according to claim 7, characterized in that The determining of the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution based on the prior distribution parameters and the return conversion number information includes: Determine the multinomial distribution according to the return conversion number information; Determine the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution; Determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula based on the prior distribution parameters and the return conversion number information.

9. The method according to claim 8, characterized in that, The determining of the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution includes: Determine a second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution; Determine the Bayesian estimation formula according to the second Dirichlet distribution.

10. The method according to claim 8, characterized in that, The determining of the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula based on the prior distribution parameters and the return conversion number information includes: Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number is one of the return conversion numbers in the return conversion number information; Based on the N prior parameters and the target return conversion number, the target return ratio corresponding to the target advertisement in the target time period is calculated through the Bayesian estimation formula.

11. The method according to claim 1, characterized in that, The determining the ranking result of the target advertisement among the P advertisements according to the estimated conversion number includes: Determining a future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number; the future cost achievement adjustment coefficient is a coefficient used to adjust the bid in the oCPA bidding ranking formula to make the final cost close to the target cost; Determining the revenue per thousand impressions corresponding to the target advertisement according to the future cost achievement adjustment coefficient corresponding to the target advertisement; Obtaining the revenue per thousand impressions corresponding to each advertisement to be delivered among the P advertisements, where the revenue per thousand impressions corresponding to each advertisement to be delivered is determined according to the future cost achievement adjustment coefficient corresponding to each advertisement to be delivered; The formula for the revenue per thousand impressions is: revenue per thousand impressions = (target cost × billing ratio coefficient × future cost achievement adjustment coefficient) × (pCVR × pCVR correction coefficient) × pCTR; Wherein, the target cost represents the cost that the advertiser is willing to pay; the billing ratio coefficient represents the ratio between the actual deduction after click and the target cost; pCVR represents the estimated conversion rate; the pCVR correction coefficient is used to correct pCVR according to historical data; pCTR represents the prediction of the advertisement click-through rate; Ranking the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each advertisement to be delivered to obtain the ranking result of the target advertisement among the P advertisements.

12. The method according to claim 1, wherein The method further includes: Pushing the ranking result of the target advertisement among the P advertisements to the client; Receiving an advertisement placement selection instruction sent by the client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, where Q is an integer greater than or equal to 1 and less than or equal to P; If the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement, it is determined that the target advertisement meets the advertisement placement condition.

13. An advertising placement device, characterized in that, Includes: An acquisition module, configured to acquire the return ratio information corresponding to the P advertisements in the historical time period, and the return conversion number information corresponding to the target advertisement in the historical time period, where the P advertisements at least include the target advertisement, and the P advertisements all correspond to the same advertiser identifier, and P is an integer greater than or equal to 1; the return ratio information represents the ratio of the return conversion number to the total conversion number; the return conversion number information represents the converted number that has been returned; A determination module, configured to determine the prior distribution parameters corresponding to the prior distribution according to the return ratio information; The determination module is further configured to determine the target return ratio corresponding to the target advertisement in the target time period through the posterior distribution based on the prior distribution parameters and the return conversion number information, where the target time period belongs to a time period within the historical time period; The determining module is further configured to determine the estimated conversion number of the target advertisement according to the target return ratio and the return conversion number information; The determining module is further configured to determine the sorting result of the target advertisement among the P advertisements according to the estimated conversion number; The placing module is configured to place the target advertisement if it is determined that the target advertisement meets the advertisement placement condition according to the sorting result.

14. The device according to claim 13, characterized in that, The obtaining module is specifically configured to: Obtain the return ratio information corresponding to the P advertisements in the historical time period, where the historical time period includes N time periods, the return ratio information includes N return ratios, the return ratio has a corresponding relationship with the time period, and N is an integer greater than or equal to 1; Obtain the return conversion number information corresponding to the target advertisement in the historical time period, where the return conversion number information includes N return conversion numbers, the return conversion number has a corresponding relationship with the time period, and the N return conversion numbers; Wherein, the target advertisement is a low-frequency advertisement or a non-low-frequency advertisement, the low-frequency advertisement is an advertisement whose sum of the N return conversion numbers is less than or equal to the return conversion number threshold, and the non-low-frequency advertisement is an advertisement whose sum of the N return conversion numbers is greater than the return conversion number threshold.

15. The device according to claim 13 or 14, characterized in that, The prior distribution is the first beta distribution; the determining module is specifically configured to: Calculate the mean value and variance according to the return ratio information; Determine the first prior parameter in the prior distribution parameters according to the mean value and the variance; Determine the second prior parameter in the prior distribution parameters according to the mean value and the variance, where the second prior parameter and the first prior distribution parameter belong to the prior distribution parameters corresponding to the first beta distribution.

16. The device according to claim 15, characterized in that, The determining module is specifically configured to: Determine the binomial distribution according to the return conversion number information; Determine the Bayesian estimation formula corresponding to the posterior distribution according to the binomial distribution and the first beta distribution; Based on the prior distribution parameters and the return conversion number information, determine the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

17. The device according to claim 16, characterized in that, The determining module is specifically configured to: Determine the second beta distribution according to the binomial distribution and the first beta distribution, where the first beta distribution belongs to the prior distribution and the second beta distribution belongs to the posterior distribution; Determine the Bayesian estimation formula according to the second beta distribution.

18. The device according to claim 16, characterized in that, The determining module is specifically configured to: Obtain the target return conversion number corresponding to the target advertisement in the target time period, where the target return conversion number belongs to one of the return conversion numbers in the return conversion number information; Based on the first prior parameter, the second prior parameter and the target return conversion number, calculate the target return ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

19. The device according to claim 13 or 14, characterized in that The prior distribution is the first Dirichlet distribution; the determining module is specifically configured to: Calculate the average value and variance based on the reflux ratio information; Determine N prior parameters among the prior distribution parameters according to the average value and the variance, where the N prior parameters belong to the prior distribution parameters corresponding to the first Dirichlet distribution, and N is an integer greater than or equal to 1.

20. The device according to claim 19, characterized in that, The determining module is specifically configured to: Determine a multinomial distribution according to the reflux conversion number information; Determine the Bayesian estimation formula corresponding to the posterior distribution according to the multinomial distribution and the first Dirichlet distribution; Based on the prior distribution parameters and the reflux conversion number information, determine the target reflux ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

21. The device according to claim 20, characterized in that, The determining module is specifically configured to: Determine a second Dirichlet distribution according to the multinomial distribution and the first Dirichlet distribution, where the first Dirichlet distribution belongs to the prior distribution; Determine the Bayesian estimation formula according to the second Dirichlet distribution.

22. The device according to claim 20, characterized in that, The determining module is specifically configured to: Obtain the target reflux conversion number corresponding to the target advertisement in the target time period, where the target reflux conversion number is one of the reflux conversion numbers in the reflux conversion number information; Based on the N prior parameters and the target reflux conversion number, calculate the target reflux ratio corresponding to the target advertisement in the target time period through the Bayesian estimation formula.

23. The device according to claim 13, wherein The determining module is specifically configured to: Determine the future cost achievement adjustment coefficient corresponding to the target advertisement according to the estimated conversion number; the future cost achievement adjustment coefficient is a coefficient used to adjust the bid price in the oCPA bidding ranking formula to make the final cost close to the target cost; Determine the revenue per thousand impressions corresponding to the target advertisement according to the future cost achievement adjustment coefficient corresponding to the target advertisement; Obtain the revenue per thousand impressions corresponding to each to-be-delivered advertisement among the P advertisements, where the revenue per thousand impressions corresponding to each to-be-delivered advertisement is determined according to the future cost achievement adjustment coefficient corresponding to each to-be-delivered advertisement; The formula for the revenue per thousand impressions is: revenue per thousand impressions = (target cost × billing ratio coefficient × future cost achievement adjustment coefficient) × (pCVR × pCVR correction coefficient) × pCTR; Wherein, the target cost represents the cost that the advertiser is willing to pay; the billing ratio coefficient represents the ratio between the actual deduction after click and the target cost; pCVR represents the estimated conversion rate; the pCVR correction coefficient is used to correct pCVR according to historical data; pCTR represents the advertisement click-through rate prediction; Sort the revenue per thousand impressions corresponding to the target advertisement and the revenue per thousand impressions corresponding to each to-be-delivered advertisement to obtain the ranking result of the target advertisement among the P advertisements.

24. The device according to claim 13, characterized in that The device further includes: a push module and a receiving module; The push module is configured to push the ranking result of the target advertisement among the P advertisements to the client; The receiving module is configured to receive an advertisement placement selection instruction sent by a client, where the advertisement placement selection instruction carries at least Q advertisement identifiers, and Q is an integer greater than or equal to 1 and less than or equal to P; The determining module is further configured to determine that the target advertisement meets the advertisement placement condition if the Q advertisement identifiers include the advertisement identifier corresponding to the target advertisement.

25. A computer device, characterized in that, Comprising: A memory, a transceiver, a processor, and a bus system; Wherein, the memory is used to store programs; The processor is configured to execute the programs in the memory, and the processor is configured to execute the method according to any one of claims 1 to 12 based on the instructions in the program code; The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.

26. A computer-readable storage medium, characterized in that, Comprising instructions that, when running on a computer, cause the computer to execute the method according to any one of claims 1 to 12.

27. A computer program product including a program, characterized in that, When running on a computer, cause the computer to execute the method according to any one of claims 1 to 12.

Images