Multi-channel advertisement putting method and device and electronic equipment

Abstract

The invention discloses a multi-channel advertisement putting method and apparatus, and an electronic device. The method comprises the steps of determining a total advertisement putting budget and an advertisement conversion target; according to historical advertisement putting data of a plurality of channels, with the maximum conversion amount of the advertisement conversion target in the plurality of channels as a target, allocating the total advertisement putting budget to obtain advertisement putting budgets of the plurality of channels; and according to the advertisement putting budget of the plurality of channels, determining advertisement bids of the plurality of channels.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of advertising placement, and more specifically, to a multi-channel advertising placement method, device, and electronic device. Background Art

[0002] With the development of Internet technology, advertising placement platforms have become more intelligent and automated. More and more advertisers choose to conduct online advertising placement on the Internet. Currently, the commonly used online advertising automatic placement modes include: Optimized Cost Per Click (OCPC), Optimized Cost Per Mille (OCPM), and Optimized Cost Per Download (OCPD).

[0003] In the OCPX (such as OCPC, OCPM, OCPD, etc.) placement mode, when an advertiser creates an advertising plan, its placement parameters include the budget and the expected cost per action (CPA), etc. When the actual CPA exceeds the expected CPA by a certain margin, the advertising placement platform needs to compensate the advertiser. Therefore, during the advertising placement process, when the actual CPA is higher than a certain preset CPA threshold, the advertising placement platform will lower the bid; when the actual CPA is lower than a certain preset CPA threshold, the advertising placement platform will raise the bid, so as to realize the automatic advertising placement in the OCPX mode and ensure the advertising placement cost.

[0004] However, during the process of automatic advertising placement in the above OCPX mode, some advertisements may have problems with difficult volume running. For example, for advertisements with high bids and sparse conversions, most of the time in the OCPX mode will be in a suppressed state (i.e., the state of lowering the bid), which will lead to the phenomenon of difficult volume running. Summary of the Invention

[0005] The embodiments of the present application provide a multi-channel advertising placement method, device, and electronic device. The following introduces each aspect involved in the embodiments of the present application.

[0006] In a first aspect, a multi-channel advertising placement method is provided, including: determining the total advertising placement budget and the advertising conversion target; allocating the total advertising placement budget based on the historical advertising placement data of multiple channels, with the goal of maximizing the conversion volume of the advertising conversion target in the multiple channels, to obtain the advertising placement budgets of the multiple channels; determining the advertising bids of the multiple channels according to the advertising placement budgets of the multiple channels.

[0007] Second aspect, a multi-channel advertisement placement device is provided, including: a first determination module, configured to determine the total advertisement placement budget and the advertisement conversion target; an allocation module, configured to allocate the total advertisement placement budget based on the historical advertisement placement data of multiple channels, with the goal of maximizing the conversion volume of the advertisement conversion target among the multiple channels, to obtain the advertisement placement budgets of the multiple channels; a second determination module, configured to determine the advertisement bids of the multiple channels according to the advertisement placement budgets of the multiple channels.

[0008] Third aspect, a chip is provided, including: a processor, configured to call and run a computer program from a memory, so that a device installed with the chip executes the method described in the first aspect.

[0009] Fourth aspect, an electronic device is provided, including a processor and a memory, where the memory is configured to store a computer program, and the processor is configured to call and run the computer program stored in the memory to control the electronic device to execute the method described in the first aspect.

[0010] Fifth aspect, a computer-readable storage medium is provided, on which an executable code is stored, and when the executable code is executed, it can implement the method described in the first aspect.

[0011] An embodiment of the present application provides a multi-channel advertisement placement method, including: determining the total advertisement placement budget and the advertisement conversion target; allocating the total advertisement placement budget based on the historical advertisement placement data of multiple channels, with the goal of maximizing the conversion volume of the advertisement conversion target among the multiple channels, to obtain the advertisement placement budgets of the multiple channels; determining the advertisement bids of the multiple channels according to the advertisement placement budgets of the multiple channels. Aiming at the problem of difficult volume running in the OCPX mode, this solution allocates the advertisement placement budgets of multiple channels with the goal of maximizing the conversion volume of the advertisement conversion target, and can obtain the optimal budget allocation on each traffic channel (budget allocation with the goal of maximizing the conversion volume), so as to achieve the purpose of fast volume start and stable consumption. Description of the Drawings

[0012] Figure 1 is a schematic flowchart of a multi-channel advertisement placement method provided by an embodiment of the present application.

[0013] Figure 2 is a schematic structural diagram of a system for calculating the PID bid adjustment factor in the calculation plan dimension provided by an embodiment of the present application.

[0014] Figure 3 is a schematic structural diagram of a multi-channel intelligent bidding system provided by an embodiment of the present application.

[0015] Figure 4It is a schematic structural diagram of a multi-channel budget allocation training system provided by an embodiment of the present application.

[0016] Figure 5 It is a schematic structural diagram of a multi-channel intelligent bidding system provided by another embodiment of the present application.

[0017] Figure 6 It is a schematic structural diagram of a multi-channel advertising placement device provided by an embodiment of the present application.

[0018] Figure 7 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0020] In recent years, online advertising has gradually emerged and become the new favorite of advertisers. For most Internet companies, advertising has become one of the main revenue sources. Inside large Internet advertising platforms, such as ByteDance, Tencent, Xiaomi, etc., there are various heterogeneous segmented traffic, forming different advertising placement channels, such as search, recommendation, alliance, etc. Each advertising placement channel usually uses an independent set of bidding systems, and the differences in the bidding atmosphere further lead to the differences in advertising placement channels. For advertisers, it is very costly in terms of learning and time to understand the characteristics of each placement channel and formulate a suitable placement plan.

[0021] In response to this situation, mature advertising platforms will provide advertisers with an automated placement service. It does not require advertisers to perceive the numerous placement channels inside the platform. Advertisers only need to formulate a set of placement parameters according to their own needs, and the platform takes the optimal placement effect as the goal and automatically places advertisements on different internal traffic channels. Currently, the commonly used advertising automatic placement modes include: optimized cost per click OCPC, optimized cost per thousand (impressions) OCPM, and optimized cost per download OCPD. In the OCPX (such as OCPC, OCPM, OCPD, etc.) placement mode, when an advertiser creates an advertising plan, its placement parameters include the budget and the expected cost per action CPA, etc. When the actual CPA exceeds the expected CPA by a certain margin, the advertising placement platform needs to compensate the advertiser. Therefore, during the advertising placement process, when the actual CPA is higher than a certain preset CPA threshold, the advertising placement platform will lower the bid; when the actual CPA is lower than a certain preset CPA threshold, the advertising placement platform will raise the bid, so as to achieve the automatic advertising placement in the OCPX mode and ensure the advertising placement cost.

[0022] However, during the automatic delivery of ads in the above-mentioned OCPX mode, some ads may have difficulty in running volume. For example, for ads with high bids and sparse conversions, the OCPX mode will be in a suppressed state (i.e., the bid is lowered) most of the time, which will lead to the phenomenon of difficulty in running volume. For another example, in addition to daily delivery of online ads, some advertisers will also use it for impulse delivery, such as raising the budget at the end of the month, holidays and other time windows, hoping to obtain more conversions than usual. Under the OCPX delivery model, the competitiveness of an ad has nothing to do with the budget. Therefore, even if advertisers increase their budgets, they still cannot achieve the goal of obtaining more conversions. Under the OCPX delivery model, the competitiveness of an ad, ecpm, can be expressed as:

[0023] ecpm=pctr*pcvr*CPA*λ

[0024] Among them, pctr is the predicted click-through rate (pctr), pcvr is the predicted conversion rate (pcvr), CPA is the cost per conversion, and λ is the advertising bid regulation factor.

[0025] It should be understood that difficulty in increasing traffic is disadvantageous to both advertisers and platforms: for advertisers, they cannot obtain more conversions; for platforms, potential revenue is lost.

[0026] In response to the above problems, the embodiments of the present application provide a multi-channel advertising method, including: determining the total advertising budget and advertising conversion targets; allocating the total advertising budget based on the historical advertising data of multiple channels, with the goal of maximizing the conversion volume of advertising conversion targets in multiple channels, to obtain advertising budgets for multiple channels; determining advertising bids for multiple channels based on the advertising budgets of multiple channels. In response to the problem of difficulty in running volume under the OCPX mode, this solution allocates advertising budgets for multiple channels with the goal of maximizing the conversion volume of advertising conversion targets, and can obtain the optimal budget allocation on each traffic channel, thereby achieving the purpose of fast volume growth and stable consumption.

[0027] It should be noted that the execution subject in the embodiments of this application can be an electronic device, which can refer to a multi-channel advertising placement device, or a terminal or a server. The terminal can refer to a mobile phone, a tablet computer, a laptop computer, a palm computer, a mobile internet device (MID), etc. The server can be an independent server 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 network (CDN), and big data and artificial intelligence platforms. This application does not make specific limitations on this.

[0028] It can be understood that the embodiments of this application can obtain and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is to use a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results of theory, methods, technologies, and application systems. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technologies mainly include several major directions such as computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0029] The following combines Figure 1 to introduce the multi-channel advertising placement method in the embodiments of this application in detail. As Figure 1 shown, the multi-channel advertising placement method 100 may include steps: S110 to S130.

[0030] In step S110, determine the total advertising placement budget and the advertising conversion target.

[0031] The total advertising placement budget may refer to the advertising budget of an advertiser within a certain period of time, such as the advertising budget of an advertiser in one day.

[0032] In this embodiment, the advertising conversion target may refer to the advertising placement conversion effect that the advertiser hopes to obtain. The advertising conversion target includes but is not limited to exposure, click, registration, download, order placement, payment, binding, etc. The advertiser can set it according to needs. This application does not make specific limitations on this.

[0033] In step S120, based on the historical advertisement delivery data of multiple channels, with the goal of maximizing the conversion volume of the advertisement conversion target among the multiple channels, the total advertisement delivery budget is allocated to obtain the advertisement delivery budgets of the multiple channels.

[0034] The historical advertisement delivery data of multiple channels may refer to the historical advertisement delivery data of multiple channels on an advertisement platform within a period of time (such as one week or one day, etc.). This historical advertisement delivery data may include exposure rate, exposure volume, click-through rate, click volume, download rate, download volume, etc., and the present application does not make specific limitations thereto.

[0035] The maximized conversion volume of the advertisement conversion target among the above-mentioned multiple channels may refer to the goal to be achieved after the allocation of the advertisement budget. That is to say, the total advertisement delivery budget may be allocated to multiple channels in different proportions, and there may be one or more budget allocation proportions that enable the advertisement conversion target among the multiple channels to reach the maximized conversion volume. This one or more budget allocation proportions are our target allocation proportions.

[0036] As an example, when the advertisement conversion target is registration, the total advertisement delivery budget may be allocated with the goal of maximizing the conversion volume of registration. Of course, it is also possible to allocate the total advertisement delivery budget with the goal of maximizing the conversion volume of advertisement conversion targets such as placing an order or making a payment, and the present application does not make specific limitations thereto.

[0037] It should be understood that by allocating the total advertisement delivery budget with the goal of maximizing the conversion volume of the advertisement conversion target among multiple channels, the advertisement delivery budget that can be allocated to each channel among the multiple channels can be obtained.

[0038] The maximized conversion volume in the embodiments of the present application may refer to the maximized conversion number of the advertisement.

[0039] In some embodiments, the historical advertisement delivery data of multiple channels includes the exposure rate of each advertisement request, the click-through rate of each advertisement request, the conversion rate of each advertisement request, and the payment for obtaining exposure for each advertisement request among the multiple channels. Allocating the total advertisement delivery budget may include: determining a first objective function for the maximized conversion volume of the advertisement conversion target among multiple channels according to the exposure rate, click-through rate, and conversion rate; determining a first constraint function of the first objective function according to the exposure rate, the payment for obtaining exposure for each advertisement request, and the total advertisement delivery budget; and allocating the total advertisement delivery budget by using linear programming according to the first objective function and the first constraint function to obtain the advertisement delivery budgets of multiple channels.

[0040] The first objective function is:

[0041]

[0042] The first constraint condition is:

[0043]

[0044] where x ij is the exposure rate of the j-th request of the advertisement on the i-th channel, ctr ij is the click-through rate of the j-th request of the advertisement on the i-th channel, cvr ij is the conversion rate of the j-th request of the advertisement on the i-th channel, wp ij is the payment for the exposure of the j-th request of the advertisement on the i-th channel, z i is the budget allocation ratio of the i-th channel, B is the total advertising budget, M i is the total number of advertisement requests of the i-th channel, N is the total number of channels, and both i and j are positive integers.

[0045] It should be noted that the exposure rate of the j-th request of the advertisement on the i-th channel may refer to whether the j-th request of the advertisement on the i-th channel is exposed. This exposure rate can be 0 or 1. As an example, if the exposure rate of the j-th request of the advertisement on the i-th channel is 0, it means that the j-th request of the advertisement on the i-th channel is not exposed; if the exposure rate of the j-th request of the advertisement on the i-th channel is 1, it means that the j-th request of the advertisement on the i-th channel is exposed.

[0046] It should be understood that after obtaining the budget allocation ratio z i of each channel, multiplying z i by the total advertising budget can obtain the advertising budget of each channel.

[0047] In step S130, according to the advertising budgets of multiple channels, determine the advertising bids of multiple channels.

[0048] Determining the advertising bids of multiple channels may refer to determining the advertising bids of each channel among multiple channels according to the advertising budget of each in multiple channels.

[0049] It can be seen from the above description that the embodiments of the present application aim at maximizing the conversion volume of the advertising conversion target, allocate the advertising budgets of multiple channels, and can obtain the optimal budget allocation on each traffic channel (budget allocation aiming at maximizing the conversion volume), so as to achieve the purpose of fast start-up and stable consumption.

[0050] In order to fully exploit the traffic value of each channel, the present application also adjusts the advertising bids of each channel so that each channel can have a better conversion volume, so that the effect ceiling of each channel is higher.

[0051] In some embodiments, the advertising bid for the first channel can be determined based on the historical advertising placement data of the first channel and the advertising placement budget of the first channel, with the goal of maximizing the conversion volume of the advertising conversion target in the first channel.

[0052] The advertising bid for the first channel can refer to the advertising bid for the first time period (daily, hourly, or second-level). This application does not make specific restrictions on the length of the first time period and can be set according to requirements. As an example, the advertising bid for the first channel can refer to the advertising bid for the first channel within a day or the advertising bid for the first channel within an hour. It should be understood that the advertising bid for the first channel needs to be updated every first time period.

[0053] The first channel can refer to any one of the above-mentioned multiple channels. The advertising placement budget for the first channel can refer to the budget that can be allocated to the first channel after the total advertising placement budget is allocated in step S120.

[0054] The historical advertising placement data of the first channel includes the exposure rate of each advertising request in the first channel, the click-through rate of each advertising request in the first channel, the conversion rate of each advertising request in the first channel, and the payment for each advertising request that obtains exposure in the first channel, etc. In some embodiments, determining the advertising bid for the first channel can include: determining a second objective function for maximizing the conversion volume of the advertising conversion target in the first channel according to the exposure rate of the advertisement in the first channel, the click-through rate of the advertisement in the first channel, and the conversion rate of the advertisement in the first channel; determining a second constraint function of the second objective function according to the exposure rate of the advertisement in the first channel, the payment for each time the advertisement obtains exposure in the first channel, and the advertising placement budget of the first channel; and determining the advertising bid for the first channel according to the second objective function and the second constraint function.

[0055] In some embodiments, determining the advertising bid for the first channel can include: converting the second objective function and the second constraint function into corresponding Lagrangian functions; then, performing a derivative operation on the Lagrangian function to obtain the derivative function of the Lagrangian function; and then, determining the advertising bid for the first channel according to the derivative function by using the complementary slackness theorem.

[0056] In some embodiments, determining the advertising bid for the first channel can include: determining the advertising bid for the first channel according to the virtual single conversion cost of the first channel and the first target bid adjustment factor of the first channel.

[0057] As an example, the advertising bid function of the first channel can be determined according to the derivative function of the Lagrangian function by using the complementary slackness theorem; according to the virtual single-conversion cost of the first channel and the first target bid regulation factor of the first channel, the advertising bid of the first channel can be determined by using the advertising bid function. It should be understood that before determining the advertising bid of the first channel, it also includes: determining the virtual single-conversion cost of the first channel and the first target bid regulation factor of the first channel.

[0058] Exemplarily, the second objective function is:

[0059]

[0060] The second constraint condition is:

[0061]

[0062] The Lagrangian function L(x i ) is:

[0063]

[0064] The derivative function is:

[0065] ―ctr i ·cvr i +α·wp i ―β i +γ i =0 (6)

[0066] Furthermore, it can be known from the complementary slackness theorem that:

[0067]

[0068] When x i =1, there is:

[0069]

[0070] Thus, the advertising bid function can be obtained as:

[0071]

[0072]

[0073] Among them, x i is the exposure rate of the advertisement on the i-th channel, ctr i is the click-through rate of the advertisement on the i-th channel, cvr i is the conversion rate of the advertisement on the i-th channel, wp i is the payment for the advertisement to obtain exposure on the i-th channel, B iis the advertising budget for the i-th channel, α, β i , γ i are the Lagrangian transformation coefficients of the second constraint condition, vCPA is the virtual cost per acquisition for the i-th channel, nobidFactor is the first target bid adjustment factor for the i-th channel, and bid i is the advertising bid for the advertisement in the i-th channel.

[0074] In order to fully exploit the traffic value of each channel, the present application also adjusts the hourly advertising bid for each channel, so that each channel can have better conversion volume, and thus the ceiling of the effect of each channel is higher. The hourly level can refer to a period of time with an hour as the adjustment unit. For example, the advertising bid can be adjusted once an hour, or once every two hours. The present application does not make specific restrictions on this. The method for determining the virtual cost per acquisition of the first channel and the first target bid adjustment factor at the hourly level is illustrated by way of example below.

[0075] In some embodiments, the virtual cost per acquisition of the first channel can be determined according to the historical advertising placement data of different dimensions in the first channel. For example, the virtual cost per acquisition in the first channel is any one of the following conversion costs: the average cost per acquisition of the advertising conversion type dimension and the advertising ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application classification ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension in the first time period in the first channel. Among them, the first time period can refer to one day or multiple days, and of course it can also refer to one week or multiple weeks. The present application does not make specific restrictions on this.

[0076] It should be noted that the advertising conversion type dimension can refer to one advertising conversion type (i.e., advertising conversion target) among multiple advertising conversion types. The advertising conversion types include but are not limited to exposure, click, registration, download, order placement, payment, binding, etc. The advertiser can set according to needs. The present application does not make specific restrictions on this. The advertising ID dimension can refer to one advertisement among multiple advertisements, and each advertisement has a unique identification ID. The application ID dimension can refer to one application among multiple applications, and each application has a unique identification ID. Each application can include one or more advertisements. The application classification ID dimension can refer to one application classification among multiple application classifications, and each application classification has a unique identification ID. Each application classification includes one or more applications.

[0077] In some embodiments, the virtual cost per single conversion in the first channel may be the cost per single conversion at the daily level. Exemplarily, taking the ad conversion type as download, the application ID as the Douyin application, and the first time period as 7 days, the cost per single conversion of the ad conversion type dimension and the application ID dimension in the first channel within the first time period may refer to the average cost per single conversion of the first channel within 7 days in the dimensions of the Douyin application and ad download.

[0078] It can be understood that the range sizes of the above four dimensions are: ad conversion type dimension > ad conversion type dimension and application classification ID dimension > ad conversion type dimension and application ID dimension > ad conversion type dimension and ad ID dimension.

[0079] As the range of data dimensions shrinks, the granularity of ad placement data becomes higher. The higher the granularity of the data, the more precise the ad bid control, and thus the more effective it is to explore the effects of traffic channels. Therefore, when an advertiser creates a new maximize conversion ad plan, they can first find the average conversion cost of the conversion type * ad ID dimension as the virtual CPA; if not found, then continue to find the average conversion cost of the conversion type * application ID dimension, and so on.

[0080] In some embodiments, determining the first target bid adjustment factor for the first channel includes: determining the expected allocation budget within the target time period according to the conversion volume distribution of the ad conversion target in the first channel; determining the expected consumption budget within the target time period according to the expected allocation budget within the target time period; and determining the first target bid adjustment factor for the first channel according to the expected consumption budget within the target time period and the actual consumption budget within the target time period. The target time period may be a time period at the daily level. For example, if the target time period is one day, then the first target bid adjustment factor is a bid adjustment factor at the daily level, and only one bid is used per day. The target time period may also be a time period at the hourly level. For example, the target time period is one hour or two hours, etc., then the first target bid adjustment factor is a bid adjustment factor at the hourly level (such as the ad bid can be adjusted once per hour). Similarly, the target time period can also be a time period at the second level, indicating that the ad bid can be adjusted once every second-level target time period (such as 30 seconds), and the regulation is more precise. The following takes the hourly bid adjustment as an example for illustration.

[0081] In some embodiments, according to the historical conversion volume data of the ad conversion target in the first channel, determine the first distribution of the hourly conversion volume of the first channel in the first ad dimension and determine the second distribution of the hourly conversion volume of the first channel in the second ad dimension; according to the first distribution and the second distribution, use Bayesian inference to determine the posterior distribution of the hourly conversion volume of the first channel; and determine the expected allocation budget within the target time period according to the posterior distribution.

[0082] In some embodiments, the first advertising dimension is the advertising conversion type dimension, the first distribution is the Dirichlet distribution, the second advertising dimension is the advertising conversion type dimension and the application ID dimension, the second distribution is the multinomial distribution, the first distribution can be the prior distribution, and the prior distribution and the posterior distribution are conjugate distributions.

[0083] Exemplarily, for the hourly conversion volume distribution information of the advertising conversion type dimension, the Dirichlet distribution can be used:

[0084]

[0085] where is the proportion of the advertising conversion volume of the advertising conversion type dimension in the target time period (such as a certain hour or consecutive hours) to the total advertising conversion volume of the whole day, is the Dirichlet distribution parameter.

[0086] The hourly conversion volume information of the application ID * conversion type dimension can be used as sample information, and this sample information follows the multinomial distribution, that is where is the advertising conversion volume generated by the application ID * conversion type dimension in the target time period (such as a certain hour or consecutive hours).

[0087] That is to say, based on the Dirichlet distribution (prior distribution) and the multinomial distribution (sample information), the posterior distribution of the hourly conversion volume of the first channel can be determined using Bayesian inference. Among them, the prior distribution and the posterior distribution are conjugate distributions, that is, the posterior distribution is also the Dirichlet distribution.

[0088] Furthermore, Bayesian posterior inference can be used to solve the optimal budget allocation ratio. For example, the moment estimation method can be used to solve the Dirichlet distribution parameter of the multinomial distribution parameters

[0089] where a 1 represents the advertising conversion volume generated by the advertising conversion type dimension from 1:00 to 2:00, a 2 represents the advertising conversion volume generated by the advertising conversion type dimension from 2:00 to 3:00, and so on. Similarly, c 1 represents the advertising conversion volume generated by the application ID * conversion type dimension from 1:00 to 2:00, c 2 represents the advertising conversion volume generated by the application ID * conversion type dimension from 2:00 to 3:00, and so on.

[0090] For the multinomial distribution in Represents the proportion distribution of the advertising conversion volume for the application ID * conversion type dimension. For example, r 1 Represents the proportion of the advertising conversion volume generated for the application ID * conversion type dimension from 1:00 to 2:00 in the total advertising conversion volume for the whole day, r 2 Represents the proportion of the advertising conversion volume generated for the application ID * conversion type dimension from 2:00 to 3:00 in the total advertising conversion volume for the whole day, r 23 Represents the proportion of the advertising conversion volume generated for the application ID * conversion type dimension from 23:00 to 24:00 in the total advertising conversion volume for the whole day, and so on.

[0091] The Bayesian posterior estimate is:

[0092]

[0093] Among them, r i Represents the proportion of the advertising conversion volume in the first channel. Therefore, r i Can be used as the expected budget proportion to be allocated for the i-th hour in the first channel, so as to make the effect ceiling of the first channel higher.

[0094] It can be understood that if the budget consumption curve is close to the advertising conversion volume distribution curve, the traffic can be more reasonably allocated. The advertiser gets more conversion volume, and at the same time, the overall advertising conversion volume of the platform will also be more. That is to say, the expected consumption budget within the target time period can be equal to the expected allocation budget within the target time period.

[0095] In some embodiments, the expected consumption budget within the target time period includes the expected consumption budget in the plan dimension within the target time period and the expected consumption budget in the account dimension within the target time period. According to the expected consumption budget within the target time period and the actual consumption budget within the target time period, determine the first target bid adjustment factor for the first channel, including: determining the bid adjustment factor for the plan dimension of the first channel according to the expected consumption budget in the plan dimension within the target time period and the actual consumption budget in the plan dimension within the target time period; determining the bid adjustment factor for the account dimension of the first channel according to the expected consumption budget in the account dimension within the target time period and the actual consumption budget in the account dimension within the target time period; determining the first target bid adjustment factor for the first channel, and the first target bid adjustment factor is the smallest bid adjustment factor among the bid adjustment factor for the plan dimension of the first channel and the bid adjustment factor for the account dimension of the first channel.

[0096] In some embodiments, the bid adjustment factor for the plan dimension of the first channel and the bid adjustment factor for the account dimension of the first channel are PID bid adjustment factors.

[0097] Exemplarily, see Figure 2, the calculation process of the PID bid adjustment factor for the first channel plan dimension is as follows:

[0098] Step (1.1), calculate the proportional error term (P-term error). Calculate the proportion of the proportional error according to the actual consumption budget and the expected consumption budget as:

[0099]

[0100] where realCost is the actual consumption budget, expectCost is the expected consumption budget, and the expected consumption budget can be r i *B i , and smoothValue is the smoothing term.

[0101] Step (1.2), calculate the integral error term (I-term error). It is the cumulative term of the proportional error term, and the integral error integral is:

[0102] integral = ∫proportion (14)

[0103] Step (1.3), calculate the differential error term (D-term error). The differential error term is the ratio of the current proportional error term and the previous proportional error term, and the differential error ifferential is:

[0104]

[0105] where m is a positive integer.

[0106] Step (1.4), generate the PID bid adjustment factor for the first channel plan dimension, and this PID bid adjustment factor λ J .

[0107] λ J = proportion kp × integral ki × differential kd (16)

[0108] where kp, ki, and kd are the control parameters of the proportional error term, the integral error term, and the differential error term respectively. These parameters can be set manually or obtained through model learning.

[0109] It should be noted that the PID bid adjustment factor λ H for the first channel account dimension is similar to the calculation process of the PID bid adjustment factor for the first channel plan dimension. The difference is that the PID bid adjustment factor λ HIt uses the advertising placement data of the account dimension (such as the actual consumption budget of the account dimension and the expected consumption budget of the account dimension), while the PID bid adjustment factor uses the advertising placement data of the plan dimension (such as the actual consumption budget of the plan dimension and the expected consumption budget of the plan dimension), which will not be elaborated here.

[0110] In some embodiments, to ensure the robustness of the PID bid adjustment factor, a basic bid adjustment factor is determined. The basic bid adjustment factor can be an offline bid adjustment factor determined according to the historical advertising placement data of the first channel.

[0111] The determination method of the basic bid adjustment factor is as follows: According to the hourly advertising click data of the first channel, a price adjustment dictionary for the target time period is determined. The price adjustment dictionary includes multiple advertising click data arranged in descending order of click cost performance. The multiple advertising click data includes multiple quantile data obtained by segmenting the multiple advertising click data. Each quantile data in the multiple quantile data includes a quantile serial number, a click cost performance index of the quantile, and a cumulative click payment of the quantile. The cumulative click payment of the quantile is the cumulative payment from the click payment of the first advertising click data in the multiple advertising click data to the click payment of the current quantile data; According to the expected allocation budget of the target time period and the price adjustment dictionary of the target time period, the basic bid adjustment factor of the first channel is determined.

[0112] In some embodiments, determining the basic bid adjustment factor of the first channel includes: comparing the expected allocation budget of the target time period with the cumulative click payments of multiple quantile data in descending order of click cost performance to determine the target quantile data. The cumulative click payment of the target quantile data is greater than or equal to the expected allocation budget of the target time period, and the cumulative click payment of the previous quantile data of the target quantile data is less than the expected allocation budget of the target time period; According to the click cost performance index of the target quantile data, the basic bid adjustment factor of the first channel is determined. The basic bid adjustment factor of the first channel is the click cost performance index of the target quantile data.

[0113] Furthermore, by combining the bid adjustment factor of the plan dimension and the bid adjustment factor of the account dimension of the channel, the first target bid adjustment factor can be determined. For example, determining the first target bid adjustment factor of the first channel includes: selecting the bid adjustment factor with the smallest value among the bid adjustment factor of the plan dimension of the first channel and the bid adjustment factor of the account dimension of the first channel as the second target bid adjustment factor; According to the second target bid adjustment factor and the basic bid adjustment factor, the first target bid adjustment factor of the first channel is determined.

[0114] In some embodiments, according to the base bid adjustment factor and the bid adjustment deviation coefficient, determine the bid adjustment range of the second target bid adjustment factor; if the second target bid adjustment factor is within the bid adjustment range, determine the first target bid adjustment factor as the second target bid adjustment factor; if the second target bid adjustment factor exceeds the bid adjustment range, determine the first target bid adjustment factor as the boundary value of the bid adjustment range.

[0115] To deepen the understanding of the method for determining the first target bid adjustment factor in the embodiments of the present application, the following will be illustrated with examples. Specifically, it may include the following steps:

[0116] Step (2.1), determine the price adjustment dictionary. Determining the price adjustment dictionary may include the following steps:

[0117] Step 1), data preprocessing.

[0118] Obtain the historical advertising placement data of the first channel, then aggregate the click data according to data dimensions such as accounts and plans, and then further divide the data at the plan dimension into hourly or daily granularity. It can be understood that an account of an advertiser on the platform may include multiple advertising plans. An advertising plan may include dimensions such as application ID or conversion type.

[0119] For example, the calculation of the click data at the plan dimension may include: first aggregate the click data according to the application ID dimension and the conversion type dimension respectively, and then divide the aggregated data into hourly granularity.

[0120] Step 2), determine the virtual cost per acquisition (vCPA).

[0121] For example, the virtual cost per acquisition (vCPA) of the first channel may be any one of the following conversion costs: the average cost per acquisition of the advertising conversion type dimension and the advertising ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application classification ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension in the first time period in the first channel.

[0122] Step 3), calculate the cost performance index λ of each click in the target time period at the plan dimension C and sort all the click data (i.e., advertising click data) in the target time period in descending order according to the cost performance index. The cost performance index λ of each click C is:

[0123] λ C = pcvr * vCPA / price d(17)

[0124] Among them, λ C is the actual deduction fee for each click, pcvr is the estimated conversion rate, and vCPA is the virtual cost per single conversion.

[0125] Step 4), for the sorted click data, X (such as 99) quantiles can be equally divided. Among them, each quantile is a click data within the target time period. Each quantile data includes the arrangement serial number of this quantile, the cost performance index λ C , and the cumulative click payment price L , price L can refer to the cumulative payment from the click payment of the first advertisement click data in multiple advertisement click data within the target time period to the click payment of the current quantile data.

[0126] Step 5), obtain an adjustment price dictionary (index, λ C , price L ). Among them, index is the quantile serial number; the cost performance index λ C can also be called the basic bid adjustment factor; price L is the cumulative click payment.

[0127] Step (2.2), determine the expected allocation budget within the target time period (such as one hour).

[0128] From formula (10), r i can be determined. Further, the proportion curHourRatio of r i occupying the remaining expected budget ratio is:

[0129]

[0130] resBudget is the remaining available allocation budget. Further, determine the total budget useBudget available for allocation to the target time period as:

[0131] useBudget = resBudget + cost i (19)

[0132] Further, the expected allocation budget curHourBudget for the current target time period can be determined as:

[0133] curHourBudget = useBudget * curHourRatio (20)

[0134] Step (2.3), determine the basic bid adjustment factor of the first channel.

[0135] Arrange the expected allocated budget curHourBudget for the target time period in descending order of click cost performance, and compare it with the cumulative click payment price at multiple quantile data points, respectively L to determine the target quantile data, where the cumulative click payment price at the quantile of the target quantile data L is greater than or equal to the expected allocated budget curHourBudget for the target time period, and the cumulative click payment price at the previous quantile data point of the target quantile data L is less than the expected allocated budget curHourBudget for the target time period.

[0136] Furthermore, determine that the basic bid adjustment factor for the first channel is the cost performance index λ of the target quantile C .

[0137] Step (2.4), determine the first target bid adjustment factor.

[0138] First, select the bid adjustment factor λ for the bid adjustment dimension of the first channel plan J and the bid adjustment factor λ for the bid adjustment dimension of the first channel account H and use the bid adjustment factor with the smallest value as the second target bid adjustment factor, that is, the second target bid adjustment factor λ m2 is:[[]]

[0139] λ m2 = min(λ H , λ J ) (21)

[0140] Then, according to the basic bid adjustment factor λ C and the bid adjustment deviation coefficient, determine the bid adjustment range of the second target bid adjustment factor. The bid adjustment deviation coefficient includes the minimum deviation coefficient minParam and the maximum deviation coefficient maxParam. The bid adjustment range of the second target bid adjustment factor λ m2 is:[[]]

[0141] λ m2 ∈ [min(λ m2 , λ C * maxParam), max(λ m2 , λ C * minParam)] (22)

[0142] Among them, minParam and maxParam can be set according to requirements.

[0143] Finally, determine the first target bid adjustment factor λ m1It should be understood that the nobidFactor mentioned above is the first target bid adjustment factor λ m1 .

[0144] If the second target bid adjustment factor λ m2 is within the above bid adjustment range, then determine the first target bid adjustment factor λ m1 as the second target bid adjustment factor λ m2 ; if the second target bid adjustment factor λ m2 exceeds the bid adjustment range, then determine the first target bid adjustment factor λ m1 as the boundary value of the bid adjustment range.

[0145] It can be understood that first, take the minimum value of the bid adjustment factor λ J in the plan dimension and the bid adjustment factor λ H in the account dimension. The reason is that from a business perspective, when there is no balance in the account, although the plan budget has not been exhausted, the advertisement cannot be played. Therefore, to maximize the conversion volume under the budget constraint, both the account budget and the plan budget must be considered simultaneously.

[0146] In addition, within a certain floating range of the basic bid adjustment factor λ C , the second target bid adjustment factor λ m2 can be used as the final adjustment factor λ m1 ; when the second target bid adjustment factor λ m2 is not within this floating range, take the floating boundary as the optimal adjustment factor. The reason is that the basic bid adjustment factor λ C is the optimal adjustment factor obtained from historical data and has very important reference value. The final adjustment factor should not deviate too much from the basic bid adjustment factor λ C ; however, the daily traffic is volatile. Therefore, the PID control algorithm can be combined to perform online fine-tuning on the basis of the basic bid adjustment factor λ C .

[0147] In some embodiments, the minute-level expected consumption budget minexpectCost can also be calculated. Then, according to the minute-level expected consumption budget minexpectCost and the minute-level actual consumption budget minrealCost, use formulas (13)-(16) to determine the minute-level PID bid adjustment factor. The minute-level bid adjustment method is similar to the hour-level bid adjustment method and will not be elaborated here.

[0148] In some embodiments, in order to fully exploit the traffic value of each channel, the present application also regulates the minute-level advertising bids for each channel, so that each channel can have better conversion volume, and further make the ceiling of the effect of each channel higher. The minute-level may refer to a period of time with minutes as the regulation unit. For example, the advertising bid can be regulated once per minute, or once every 10 minutes. The present application does not make specific restrictions on this. The following gives an example of the minute-level bid regulation.

[0149] First, calculate the hourly allocation budget. For example, the allocation budget for the most recent two hours can be calculated (if the current hour is the first hour of the day, then calculate the allocation budget for the current hour). Similar to formula (18), the proportion hourRatio of the allocation budget for the most recent two hours is:

[0150]

[0151] useBudget 2 = rseBudget 2 + cost i + cost i - 1 (24)

[0152] hourBudget = useBudget 2 * hourRatio (25)

[0153] Among them, cost i is the consumption of the advertising plan in the i-th hour; rseBudget 2 is the remaining available allocation budget, useBudget 2 is the total budget available for allocation to the target time period, and hourBudget is the allocation budget within the most recent two hours; then, calculate the minute-level expected consumption minexpectCost according to the exposure proportion within the most recent two hours as:

[0154]

[0155] minexpectCost = hourBudget * exposeRatio (27)

[0156] Among them, expose i represents the number of exposures of the advertisement within the i-th hour. For example, it can refer to the average value of the number of exposures in the past seven days in the conversion type dimension of the advertisement within the i-th hour; expose realIndicates the number of exposures within the current minute up to the i-th hour. For example, it can refer to the average number of exposures in the past seven days in the conversion type dimension within the current minute up to the i-th hour; exposeRatio is the proportion of the number of exposures within the current minute up to the i-th hour in the total number of exposures in the i-th hour and the (i - 1)-th hour. minexpectCost is the calculated expected consumption at the minute level up to the current time in the i-th hour.

[0157] The actual consumption at the minute level is the consumption of the previous hour of the advertising plan and the real-time consumption minrealCost within the current hour up to the current minute, that is:

[0158] minrealCost = mincost real + cost i-1 (28)

[0159] Then, substitute the minute-level expected consumption minexpectCost and the minute-level actual consumption minreaICost obtained in formulas (27) and (28) into formula (13). Similar to the solutions in formulas (13)-(22) above, use the solutions in formulas (13)-(22) to achieve bid regulation at the minute level, which will not be elaborated here.

[0160] To better understand the multi-channel advertising placement method in the embodiments of the present application, the following combines Figures 3 - 5 for a more detailed example.

[0161] Figure 3 It is a diagram of an intelligent bidding system for maximizing conversions in multi-channel advertising placement. As Figure 3 shown, the multi-channel advertising placement method may include steps 3.1 to 3.4.

[0162] Step 3.1, create a plan.

[0163] The advertiser creates an advertising plan for maximizing conversions on the advertising platform according to its own marketing plan, and the placement parameters include the plan budget, conversion target, etc.

[0164] Step 3.2, budget allocation.

[0165] The platform makes full use of data such as the historical requests, exposures, click-through rates, and conversion rates of the plan, and allocates the plan budget to different traffic channels through the budget allocation method.

[0166] Step 3.3, intelligent bidding.

[0167] On each traffic channel, the platform takes the planned budget as a constraint and aims to maximize the conversion volume. It makes full use of the historical data and real-time data of the advertisements to generate the advertisement bid adjustment factor at the minute or hour level. At the same time, it outputs the virtual expected conversion cost. The adjustment factor and the expected conversion cost can be used in the full-link advertisement screening links of the advertisement system, such as recall, rough ranking, fine ranking, and mixed ranking.

[0168] Step 3.4, data recovery and utilization. Disk and process data such as advertisement requests, exposures, consumption, click-through rates, conversion rates, etc., and apply them to the budget allocation method and intelligent bidding method.

[0169] Figure 4 System diagram trained for multi-channel budget allocation, as Figure 4 shown. The multi-channel budget allocation method may include steps 4.1 to 4.4.

[0170] Step 4.1, data collection.

[0171] Collect data such as advertisement requests, exposures, click-through rates, conversion rates, exposure payments, etc. from the advertisement playback engine and persist them into the big data storage system HDFS.

[0172] Step 4.2, data processing.

[0173] Take the data collected from 0:00 to 24:00 yesterday and cluster it according to the advertisement plan ID. Preprocess the data under each advertisement plan ID, including abnormal data screening, data format alignment, etc. It should be understood that each account may include multiple advertisement plan IDs.

[0174] Step 4.3, budget allocation algorithm training.

[0175] The linear programming problem of the first objective function and the first constraint condition can be solved through the scipy.optimize.linprog package provided by python. Due to the large data volume, the calculation can be performed in a parallel manner with the advertisement plan ID dimension on the hadoop platform to obtain the optimal budget allocation ratio of each plan ID on each traffic channel, and store the result in the redis cache. This process can be executed once every day at midnight.

[0176] Step 4.4, budget application. On each specific traffic channel, take maximizing the conversion volume (maximizing the number of conversions) as the adjustment target and the consumption rate of the allocated budget as the intermediate adjustment variable to perform intelligent adjustment on the advertisement bid.

[0177] Figure 5 Schematic diagram of the system structure for multi-channel intelligent bidding, as Figure 5As shown, the multi-channel intelligent bidding method may include steps 5.1 to 5.5.

[0178] Step 5.1, data dependency. Specifically, it includes the following steps:

[0179] Step 1), data preprocessing.

[0180] Obtain the historical advertising placement data of the first channel, then aggregate the click data according to data dimensions such as accounts and plans, and then further divide the data at the plan dimension into hourly or daily granularity. It can be understood that an account of an advertiser on the platform may include multiple advertising plans. An advertising plan may include dimensions such as application ID or conversion type.

[0181] For example, the calculation of the click data at the plan dimension may include: first aggregate the click data according to the application ID dimension and the conversion type dimension respectively, and then divide the aggregated data into hourly granularity.

[0182] Step 2), determine the virtual cost per acquisition (vCPA).

[0183] For example, the hourly virtual cost per acquisition (vCPA) in the first channel may be any one of the following conversion costs: the average cost per acquisition of the advertising conversion type dimension and the advertising ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension and the application classification ID dimension in the first time period in the first channel; the average cost per acquisition of the advertising conversion type dimension in the first time period in the first channel.

[0184] Step 3), calculate the cost performance index λ of each click in the target time period at the plan dimension C , and sort all the click data (i.e., advertising click data) in the target time period in descending order according to the cost performance index λ. The cost performance index λ C See formula (15).

[0185] Step 4), for the sorted click data, X (such as 99) quantiles can be equally divided. Among them, each quantile is a click data in the target time period. The data of each quantile includes the arrangement serial number of this quantile, the cost performance index λ C , and the cumulative click payment price L , price L may refer to the cumulative payment from the click payment of the first advertising click data among multiple advertising click data in the target time period to the click payment of the current quantile data.

[0186] Step 5), obtain a price adjustment dictionary including X quantiles (index, λ C , price L ).

[0187] Step 6), store the calculated dependent data in the redis cache. λ_m1

[0188] Step 5.2, update the first target bid adjustment factor λ m1 .

[0189] Calculate the PID bid adjustment factor for the first channel plan dimension, the PID bid adjustment factor for the first channel account dimension, and the basic bid adjustment factor. Further, determine the first target bid adjustment factor based on the above three bid adjustment factors.

[0190] Further, calculate and update the first target bid adjustment factor at the minute level or hour level.

[0191] Step 5.3, the virtual single conversion cost vCPA and the first target bid adjustment factor λ m1 can be applied to key links in the entire advertising chain such as recall, rough ranking, fine ranking, and mixed ranking. The purpose is to prevent a certain link in the chain from failing to perceive the adjustment factor, resulting in excessive filtering and failure to timely complete the goal of budget consumption.

[0192] In some embodiments, the embodiments of the present application can be adapted to different application scenarios through the extended sub-module method. For the budget allocation method, the optimization problem can be directly solved by modifying the constraint conditions and the optimization goal.

[0193] For example, by modifying and adding constraint conditions, the optimal budget allocation ratio for different traffic channels can be solved under the budget and cost constraint conditions, as follows:

[0194]

[0195] By solving the above optimization problem, the optimal budget allocation ratio z for different traffic channels can be obtained i .

[0196] And so on, the intelligent bidding method can first derive the optimal bid formula by modifying the problem definition, such as the derivation of the optimal bid formula for maximizing conversions under budget and cost constraints, as follows:

[0197]

[0198] The optimal bid formula can be derived, as shown in formula (23).

[0199]

[0200] Among them, p is the dual variable of the first constraint condition and is responsible for adjusting the budget consumption; q is the dual variable of the second constraint condition and is responsible for adjusting the conversion cost. Regarding the adjustment of parameters p and q, reference can be made to the foregoing, and minute-level or hour-level adjustments can be made by combining offline simulation and real-time control to achieve the optimal overall effect.

[0201] In summary, through appropriate modification, the present invention can be applied to different business scenarios and has good scalability.

[0202] It can be seen from the above content that the intelligent bidding method for maximizing conversion in multi-channel advertising investment proposed in the embodiment of the present application is an extendable daily budget allocation algorithm framework. Moreover, the daily budget allocation algorithm proposed in the present application solves the optimal budget allocation ratio of each traffic channel through operations research optimization. On the one hand, it can be extended to other resource allocation algorithms by rewriting the constraint conditions and optimization objectives; on the other hand, by using request-level data, the ceiling of the effect is improved.

[0203] Secondly, the present application also proposes an hourly budget allocation method for maximizing conversion, which can combine the coarse-grained advertiser value curve and the fine-grained advertiser value curve at the same time to achieve the goal of maximizing the number of conversions and obtain a win-win situation for both advertisers and advertising platforms.

[0204] In addition, the present application combines the historical optimal regulation factor (basic bidding regulation factor) and real-time regulation (PID bidding regulation factor), which helps to improve the regulation effect.

[0205] The foregoing has Figures 1 - 5 been described in detail in the method embodiments of the present application. Next, the device embodiments of the present application will be Figures 6 - 7 described in detail. It should be understood that the description of the method embodiments corresponds to the description of the device embodiments. Therefore, the parts not described in detail can be referred to the previous method embodiments.

[0206] Figure 6 FIG. is a schematic structural diagram of a multi-channel advertising investment device provided in an embodiment of the present application. The multi-channel advertising investment device 600 may include: a first determination module 610, an allocation module 620, and a second determination module 630.

[0207] The first determination module 610 is configured to determine the total advertising investment budget and the advertising conversion target.

[0208] The allocation module 620 is configured to allocate the total advertising investment budget according to the historical advertising investment data of multiple channels, with the goal of maximizing the conversion volume of the advertising conversion target in the multiple channels, so as to obtain the advertising investment budgets of the multiple channels.

[0209] A second determination module 630, configured to determine the advertisement bids for the multiple channels according to the advertisement placement budgets of the multiple channels.

[0210] Optionally, the historical advertisement placement data of the multiple channels includes the exposure rate of each advertisement request in the multiple channels, the click-through rate of each advertisement request, the conversion rate of each advertisement request, and the payment for obtaining exposure for each advertisement request. The allocation module 620 is configured to: determine a first objective function of the maximum conversion volume of the advertisement conversion target in the multiple channels according to the exposure rate, the click-through rate, and the conversion rate; determine a first constraint function of the first objective function according to the exposure rate, the payment for obtaining exposure for each advertisement request, and the total advertisement placement budget; and allocate the total advertisement placement budget by using linear programming according to the first objective function and the first constraint function to obtain the advertisement placement budgets of the multiple channels.

[0211] Optionally, the first objective function is: max∑ i,j x ij ·ctr ij ·vcr ij , and the first constraint condition is: where x ij is the exposure rate of the jth request of the advertisement on the ith channel, ctr ij is the click-through rate of the jth request of the advertisement on the ith channel, cvr ij is the conversion rate of the jth request of the advertisement on the ith channel, wp ij is the payment for obtaining exposure for the jth request of the advertisement on the ith channel, z i is the budget allocation ratio of the ith channel, B is the total advertisement placement budget, M i is the total number of advertisement requests of the ith channel, and N is the total number of channels.

[0212] Optionally, the multiple channels include a first channel. The second determination module 630 is configured to: determine the advertisement bid of the first channel with the maximum conversion volume of the advertisement conversion target in the first channel as the goal according to the historical advertisement placement data of the first channel and the advertisement placement budget of the first channel.

[0213] Optionally, the historical advertising placement data of the first channel includes the exposure rate of each ad request in the first channel, the click-through rate of each ad request in the first channel, the conversion rate of each ad request in the first channel, and the payment for exposure obtained for each ad request in the first channel. The allocation module 620 is configured to: determine a second objective function of the maximum conversion volume of the ad conversion target in the first channel according to the exposure rate, the click-through rate, and the conversion rate; determine a second constraint function of the second objective function according to the exposure rate, the payment for exposure obtained for each ad request, and the advertising placement budget of the first channel; and determine the ad bid of the first channel according to the second objective function and the second constraint function.

[0214] Optionally, the determining the ad bid of the first channel according to the second objective function and the second constraint function includes: converting the second objective function and the second constraint function into corresponding Lagrangian functions; performing a derivative operation on the Lagrangian function to obtain a derivative function of the Lagrangian function; and determining the ad bid of the first channel according to the derivative function by using the complementary slackness theorem.

[0215] Optionally, the apparatus further includes: a third determination module, configured to determine the virtual single conversion cost of the first channel and the first target bid adjustment factor of the first channel before determining the ad bid of the first channel. The determining the ad bid of the first channel according to the derivative function by using the complementary slackness theorem includes: determining an ad bid function of the first channel according to the derivative function of the Lagrangian function by using the complementary slackness theorem; and determining the ad bid of the first channel by using the ad bid function according to the virtual single conversion cost of the first channel and the first target bid adjustment factor of the first channel.

[0216] Optionally, the determining the virtual single conversion cost of the first channel and the first target bid adjustment factor of the first channel includes: determining the expected allocation budget within a target time period according to the conversion volume distribution of the ad conversion target in the first channel; determining the expected consumption budget within the target time period according to the expected allocation budget within the target time period; and determining the first target bid adjustment factor of the first channel according to the expected consumption budget within the target time period and the actual consumption budget within the target time period.

[0217] Optionally, determining the expected allocation budget within a target time period according to the conversion volume distribution of the advertisement conversion target in the first channel includes: determining a first distribution of the hourly conversion volume of the first channel in a first advertisement dimension and determining a second distribution of the hourly conversion volume of the first channel in a second advertisement dimension according to the historical conversion volume data of the advertisement conversion target in the first channel; determining a posterior distribution of the hourly conversion volume of the first channel by using Bayesian inference according to the first distribution and the second distribution; and determining the expected allocation budget within the target time period according to the posterior distribution.

[0218] Optionally, the first advertisement dimension is an advertisement conversion type dimension, the first distribution is a Dirichlet distribution, the second advertisement dimension is an advertisement conversion type dimension and an application ID dimension, the second distribution is a multinomial distribution, the first distribution is a prior distribution, and the first distribution and the posterior distribution are conjugate distributions.

[0219] Optionally, the expected consumption budget within the target time period includes the expected consumption budget in the plan dimension within the target time period and the expected consumption budget in the account dimension within the target time period. Determining the first target bid adjustment factor of the first channel according to the expected consumption budget within the target time period and the actual consumption budget within the target time period includes: determining the bid adjustment factor of the first channel in the plan dimension according to the expected consumption budget in the plan dimension within the target time period and the actual consumption budget in the plan dimension within the target time period; determining the bid adjustment factor of the first channel in the account dimension according to the expected consumption budget in the account dimension within the target time period and the actual consumption budget in the account dimension within the target time period; and determining the first target bid adjustment factor of the first channel, where the first target bid adjustment factor is the bid adjustment factor with the smallest value among the bid adjustment factor of the first channel in the plan dimension and the bid adjustment factor of the first channel in the account dimension.

[0220] Optionally, the bid adjustment factor of the first channel in the plan dimension and the bid adjustment factor of the first channel in the account dimension are PID bid adjustment factors.

[0221] Optionally, the device further includes: a fourth determination module, configured to determine a price adjustment dictionary for the target time period according to the hourly advertisement click data of the first channel, where the price adjustment dictionary includes a plurality of advertisement click data arranged in descending order of click cost performance, the plurality of advertisement click data includes a plurality of quantile data for segmenting the plurality of advertisement click data, and each quantile data in the plurality of quantile data includes a quantile serial number, a click cost performance index of the quantile, and a cumulative click payment of the quantile, and the cumulative click payment of the quantile is the cumulative payment from the click payment of the first advertisement click data in the plurality of advertisement click data to the click payment of the current quantile data; determine a basic bid adjustment factor for the first channel according to the expected allocation budget for the target time period and the price adjustment dictionary for the target time period; the determining the first target bid adjustment factor for the first channel includes: selecting the bid adjustment factor with the smallest value among the bid adjustment factors of the first channel plan dimension and the bid adjustment factors of the first channel account dimension as the second target bid adjustment factor; determining the first target bid adjustment factor for the first channel according to the second target bid adjustment factor and the basic bid adjustment factor.

[0222] Optionally, the determining the basic bid adjustment factor for the first channel according to the expected allocation budget for the target time period and the price adjustment dictionary for the target time period includes: comparing the expected allocation budget for the target time period with the cumulative click payment of the plurality of quantile data in descending order of click cost performance respectively to determine target quantile data, where the cumulative click payment of the target quantile data is greater than or equal to the expected allocation budget for the target time period, and the cumulative click payment of the previous quantile data of the target quantile data is less than the expected allocation budget for the target time period; determining the basic bid adjustment factor for the first channel according to the click cost performance index of the target quantile data.

[0223] Optionally, the determining the first target bid adjustment factor for the first channel according to the second target bid adjustment factor and the basic bid adjustment factor includes: determining a bid adjustment range of the second target bid adjustment factor according to the basic bid adjustment factor and a bid adjustment deviation coefficient; if the second target bid adjustment factor is within the bid adjustment range, determining the first target bid adjustment factor as the second target bid adjustment factor; if the second target bid adjustment factor exceeds the bid adjustment range, determining the first target bid adjustment factor as the boundary value of the bid adjustment range.

[0224] Optionally, the second objective function is: max∑ i x i ·ctri ·cvr i ,

[0225] The second constraint is: ∑ i x i ·wp i ≤B i , 0 ≤ x i ≤ 1, and the Lagrangian function L(x i ) is: The derivative function is: -ctr i ·cvr i + α·wp i - β i + γ i =0, and the advertising bid function is: where x i is the exposure rate of the advertisement on the i-th channel, ctr i is the click-through rate of the advertisement on the i-th channel, cvr i is the conversion rate of the advertisement on the i-th channel, wp i is the payment for the advertisement to obtain exposure on the i-th channel, B i is the advertising placement budget for the i-th channel, α, β i , γ i are the Lagrangian transformation coefficients of the second constraint, vCPA is the virtual cost per single conversion for the i-th channel, nobidFactor is the first target bid regulation factor for the i-th channel, and bid i is the advertising bid of the advertisement on the i-th channel.

[0226] Optionally, the virtual cost per single conversion at the hourly level in the first channel is any one of the following conversion costs: the average cost per single conversion of the advertisement conversion type dimension and the advertisement ID dimension in the first time period in the first channel; the average cost per single conversion of the advertisement conversion type dimension and the application ID dimension in the first time period in the first channel; the average cost per single conversion of the advertisement conversion type dimension and the application classification ID dimension in the first time period in the first channel; the average cost per single conversion of the advertisement conversion type dimension in the first time period in the first channel.

[0227] Next, a description is given in conjunction with Figure 7 An electronic device 700 in an embodiment of the present application is introduced. The electronic device 700 can be used to implement the method described in the above method embodiment. Figure 7 The dashed box in[] is an optional item.

[0228] It should be understood that the electronic device 700 can be applied to any of the types of electronic devices mentioned above.

[0229] The electronic device 700 may include one or more processors 710. The processor 710 can support the electronic device 700 to implement the methods described in the foregoing method embodiments.

[0230] The processor 710 can be a general-purpose processor or a dedicated processor. For example, the processor can be a central processing unit (CPU). Alternatively, the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0231] The electronic device 700 may further include one or more memories 720. A program is stored on the memory 720, and the program can be executed by the processor 710 to control the electronic device 700 to execute the methods described in the foregoing method embodiments. The memory 720 can be independent of the processor 710 or integrated in the processor 710.

[0232] The electronic device 700 may further include a transceiver 730. The processor 710 can communicate with other devices through the transceiver 730. For example, the processor 710 can send and receive data with other devices through the transceiver 730.

[0233] In an embodiment of the present application, a chip is further provided, including a processor, which can be used to call and run a computer program from a memory, so that a device installed with the chip executes the methods described in the foregoing method embodiments. It can be understood that the processor can be any of the types of processors mentioned above. It can be understood that the memory can be independent of the chip or integrated in the chip.

[0234] An embodiment of the present application further provides a machine-readable storage medium for storing a program. And the program enables a computer to execute the methods in various embodiments of the present application.

[0235] An embodiment of the present application further provides a computer program product. The computer program product includes a program. The program enables a computer to execute the methods in various embodiments of the present application.

[0236] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present disclosure are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a machine-readable storage medium or transmitted from one machine-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The machine-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a Digital Video Disc (DVD)), or a semiconductor medium (such as a Solid State Disk (SSD)), etc.

[0237] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments of the present disclosure can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present disclosure.

[0238] In several embodiments provided by the present disclosure, 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. In actual implementation, there can be other division methods. 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. The indirect couplings or communication connections of devices or units can be in an electrical, mechanical, or other form.

[0239] The unit described as a separation component may or may not be physically separated. The component displayed as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed over 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.

[0240] In addition, each functional unit in various embodiments of the present disclosure may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit.

[0241] As mentioned above, the above are only specific implementation manners of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present disclosure can easily think of changes or substitutions, which should all be covered within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A multi-channel advertising placement method, characterized in that, it includes: Determine the total advertising placement budget and the advertising conversion target; According to the historical advertising placement data of multiple channels, with the goal of maximizing the conversion volume of the advertising conversion target among the multiple channels, allocate the total advertising placement budget to obtain the advertising placement budgets of the multiple channels; Determine the advertising bids of the multiple channels according to the advertising placement budgets of the multiple channels.

2. The method according to claim 1, characterized in that, the historical advertising placement data of the multiple channels includes the exposure rate of each advertising request, the click-through rate of each advertising request, the conversion rate of each advertising request, and the payment for obtaining exposure for each advertising request. According to the historical advertising placement data of multiple channels, with the goal of maximizing the conversion volume of the advertising conversion target among the multiple channels, allocate the total advertising placement budget to obtain the advertising placement budgets of the multiple channels, including: Determine the first objective function for maximizing the conversion volume of the advertising conversion target among the multiple channels according to the exposure rate, the click-through rate, and the conversion rate; Determine the first constraint function of the first objective function according to the exposure rate, the payment for obtaining exposure for each advertising request, and the total advertising placement budget; Allocate the total advertising placement budget by using linear programming according to the first objective function and the first constraint function to obtain the advertising placement budgets of the multiple channels.

3. The method according to claim 1, characterized in that, the multiple channels include a first channel. Determining the advertising bids of the multiple channels according to the advertising placement budgets of the multiple channels includes: Determine the advertising bid of the first channel with the goal of maximizing the conversion volume of the advertising conversion target in the first channel according to the historical advertising placement data of the first channel and the advertising placement budget of the first channel.

4. The method according to claim 3, characterized in that, the historical advertising placement data of the first channel includes the exposure rate of each advertising request in the first channel, the click-through rate of each advertising request in the first channel, the conversion rate of each advertising request in the first channel, and the payment for obtaining exposure for each advertising request in the first channel. Determining the advertising bid of the first channel with the goal of maximizing the conversion volume of the advertising conversion target in the first channel according to the historical advertising placement data of the first channel and the advertising placement budget of the first channel includes: Determine the second objective function for maximizing the conversion volume of the advertising conversion target in the first channel according to the exposure rate, the click-through rate, and the conversion rate; Determine the second constraint function of the second objective function according to the exposure rate, the payment for obtaining exposure for each advertising request, and the advertising placement budget of the first channel; Determine the advertising bid of the first channel according to the second objective function and the second constraint function.

5. The method according to claim 4, characterized in that, Before determining the advertising bid of the first channel, the method further includes: Determining an expected allocation budget within a target time period according to the conversion volume distribution of the advertising conversion target in the first channel; Determining an expected consumption budget within the target time period according to the expected allocation budget within the target time period; Determining a first target bid adjustment factor of the first channel according to the expected consumption budget within the target time period and the actual consumption budget within the target time period.

6. The method according to claim 5, wherein, the determining an expected allocation budget within a target time period according to the conversion volume distribution of the advertising conversion target in the first channel includes: Determining a first distribution of the hourly conversion volume of the first channel in the first advertising dimension and a second distribution of the hourly conversion volume of the first channel in the second advertising dimension according to the historical conversion volume data of the advertising conversion target in the first channel; Determining a posterior distribution of the hourly conversion volume of the first channel according to the first distribution and the second distribution; Determining an expected allocation budget within the target time period according to the posterior distribution.

7. The method according to claim 5, wherein, the expected consumption budget within the target time period includes the expected consumption budget in the plan dimension within the target time period and the expected consumption budget in the account dimension within the target time period, and the determining a first target bid adjustment factor of the first channel according to the expected consumption budget within the target time period and the actual consumption budget within the target time period includes: Determining a bid adjustment factor in the plan dimension of the first channel according to the expected consumption budget in the plan dimension within the target time period and the actual consumption budget in the plan dimension within the target time period; Determining a bid adjustment factor in the account dimension of the first channel according to the expected consumption budget in the account dimension within the target time period and the actual consumption budget in the account dimension within the target time period; Determining the first target bid adjustment factor of the first channel, where the first target bid adjustment factor is the bid adjustment factor with the smallest value among the bid adjustment factor in the plan dimension of the first channel and the bid adjustment factor in the account dimension of the first channel.

8. The method according to claim 5, wherein, the determining the advertising bid of the first channel includes: Determining the advertising bid of the first channel according to the virtual single conversion cost of the first channel and the first target bid adjustment factor; The virtual single conversion cost of the first channel is any one of the following conversion costs: The average single conversion cost of the advertising conversion type dimension and the advertising ID dimension in the first channel within the first time period; The average single conversion cost of the advertising conversion type dimension and the application ID dimension in the first channel within the first time period; The average single conversion cost of the advertising conversion type dimension and the application classification ID dimension in the first channel within the first time period; The average single conversion cost of the advertising conversion type dimension in the first channel within the first time period.

9. An advertising placement device for multiple channels It is characterized in that including: A first determination module, configured to determine the total advertising budget and the advertising conversion target; An allocation module, configured to allocate the total advertising budget based on the historical advertising data of multiple channels, with the goal of maximizing the conversion volume of the advertising conversion target in the multiple channels, to obtain the advertising budgets of the multiple channels; A second determination module, configured to determine the advertising bids of the multiple channels according to the advertising budgets of the multiple channels.

10. An electronic device It is characterized in that including: A processor and a memory, the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory to control the electronic device to execute the method according to any one of claims 1-8.

Images