Method, apparatus, electronic device, and storage medium for delivering recommendation information

Abstract

An embodiment of the present application provides a method, device, electronic device, and storage medium for delivering recommendation information, which relates to the technical field of data processing. The method includes: receiving a recommendation information pulling request from a terminal, obtaining at least one recommendation information corresponding to the recommendation information pulling request according to the recommendation information pulling request, determining a shallow conversion behavior parameter and a deep conversion behavior parameter of the recommendation information, and obtaining an estimated resource acquisition amount of the recommendation information when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter; determining a target recommendation information according to the estimated resource acquisition amounts of multiple recommendation information, and returning the data corresponding to the target recommendation information to the terminal so that the terminal displays the target recommendation information. The embodiment of the present application can make the estimated resource acquisition amount of the recommendation information more accurate, reduce the occurrence probability of frequent price adjustment, and improve the delivery efficiency of the recommender.

Description

Technical Field

[0001] The present application relates to the technical field of data processing. Specifically, the present application relates to a method, apparatus, electronic device, and storage medium for delivering recommendation information. Background Art

[0002] The sorting of recommendation information is a way to display the opportunity of recommendation information in the display area of sales recommendation information often used by traffic media.

[0003] The existing recommendation information sorting mechanism requires the recommender to determine the preset cost within a delivery period by himself. After winning the opportunity to display the recommendation information, the recommender adjusts the preset cost for the next delivery period by analyzing the effect of the recommendation information within the delivery period. However, this method of adjusting the preset cost is not flexible and accurate enough, and it is easy to cause waste of recommendation information positions. Summary of the Invention

[0004] Embodiments of the present invention provide a method, apparatus, electronic device, and storage medium for delivering recommendation information that overcome the above problems or at least partially solve the above problems.

[0005] In a first aspect, a method for delivering recommendation information is provided. The method includes:

[0006] Receiving a recommendation information pull request from a terminal, and obtaining at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request;

[0007] For each of the at least one recommendation information, determining a shallow conversion behavior parameter and a deep conversion behavior parameter for each recommendation information, and obtaining an estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter;

[0008] Determining a target recommendation information from the at least one recommendation information according to the estimated resource acquisition amounts of the multiple recommendation information, and returning the data corresponding to the target recommendation information to the terminal so that the terminal displays the target recommendation information;

[0009] Wherein, the shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior, and the deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the degree of interaction between the user and the recommendation information represented by the shallow conversion behavior is lower than that of the deep conversion behavior;

[0010] The recommendation information achievement rate is used to represent the ratio of the number of recommendation information that meets the preset requirements to the number of all recommendation information obtained in advance, or the ratio of the resource occupancy of the recommendation information that meets the preset requirements to the resource consumption of all recommendation information obtained in advance;

[0011] The recommended information meeting the preset requirements satisfies that the difference between the actual cost and the preset cost of the recommended information is within the threshold.

[0012] In a possible implementation, the shallow behavior conversion parameters include at least one of the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate; the shallow conversion behaviors include at least one of exposure, click, and activation.

[0013] The deep conversion behavior parameters include at least one of the target deep conversion rate, the actual deep conversion rate, and the estimated deep conversion rate; the deep conversion behaviors include at least one of the next-day retention, three-day retention, and seven-day retention.

[0014] In a possible implementation, obtaining the estimated resource acquisition amount of the recommended information when the recommended information achievement rate is within the preset range according to the shallow conversion behavior parameters and the deep conversion behavior parameters includes:

[0015] Calculating the shallow target resource acquisition amount according to the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate of the recommended information.

[0016] Determining the floating ratio of the shallow target resource acquisition amount when the recommended information achievement rate is within the preset range according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate.

[0017] Calculating the estimated resource acquisition amount of the recommended information according to the shallow target resource acquisition amount and the floating ratio.

[0018] In a possible implementation, determining the floating ratio of the shallow target resource acquisition amount when the recommended information achievement rate of the recommended information meets the preset conditions according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate includes:

[0019] Obtaining the first quotient value of the first ratio and the first calibration coefficient, and obtaining the first coefficient according to the difference between the first quotient value and the preset reference value.

[0020] Obtaining the second quotient value of the second ratio and the second calibration coefficient, determining the difference between the second quotient value and the preset reference value, and obtaining the floating value of the second coefficient in the current cycle compared with the second coefficient in the previous cycle according to the product of the difference and the preset step size.

[0021] Obtaining the second coefficient in the previous cycle, and combining the floating value to obtain the second coefficient in the current cycle.

[0022] Obtaining the floating ratio according to the first coefficient and the second coefficient in the current cycle.

[0023] In a possible implementation, before obtaining the second quotient value of the second ratio and the second calibration coefficient, it further includes:

[0024] According to the magnitude relationship between the first coefficient and the preset threshold, determine the magnitude relationship between the preset cost of the recommended recommendation information and the initial preset cost of the recommender's recommendation information;

[0025] If the preset cost of the recommended recommendation information is higher than the initial preset cost of the recommender's recommendation information, send a cost increase notice to the recommender corresponding to the recommendation information;

[0026] Obtaining the second product result of the second ratio and the second calibration coefficient includes:

[0027] If the cancellation recommendation instruction sent by the recommender according to the cost increase notice is not received, obtain the second quotient value of the second ratio and the second calibration coefficient.

[0028] In a possible implementation, obtaining the first calibration coefficient includes:

[0029] Obtain the actual cost per action of the recommendation information in the current period;

[0030] According to the deviation between the actual cost per action of the recommendation information in the current period and the target cost per action, obtain the first calibration coefficient.

[0031] In a possible implementation, the actual deep conversion rate includes:

[0032] Obtain the estimated deep conversion rate of the recommendation information in the preset time window as the first parameter;

[0033] Obtain the deep conversion number and activation return ratio of the recommendation information in the preset time window, and obtain the second parameter according to the conversion number and activation return ratio of the recommendation information in the preset time window;

[0034] Calculate the actual deep conversion rate according to the first parameter and the second parameter.

[0035] In a second aspect, a recommendation information delivery device is provided, including:

[0036] A request acquisition module, configured to receive a recommendation information pull request from a terminal, and obtain at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request;

[0037] An estimated resource acquisition amount module, configured to, for each of at least one recommendation information, determine the shallow conversion behavior parameter and the deep conversion behavior parameter of each recommendation information, and obtain the estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter;

[0038] An information return module, configured to determine a target recommendation information from at least one recommendation information according to the estimated resource acquisition amounts of multiple recommendation informations, and return the data corresponding to the target recommendation information to a terminal, so that the terminal displays the target recommendation information;

[0039] Wherein, the shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior, and the deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the degree of interaction between the user and the recommendation information represented by the shallow conversion behavior is lower than that of the deep conversion behavior;

[0040] The recommendation information achievement rate is used to represent the ratio between the number of recommendation informations meeting the preset requirements and the number of all pre-acquired recommendation informations, or the ratio between the resource occupancy of the recommendation informations meeting the preset requirements and the resource consumption of all pre-acquired recommendation informations;

[0041] The recommendation information meeting the preset requirements satisfies that the difference between the actual cost and the preset cost of the recommendation information is within the threshold.

[0042] In a possible implementation manner, the shallow behavior conversion parameter includes at least one of the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate; the shallow conversion behavior includes at least one of exposure, click, and activation;

[0043] The deep conversion behavior parameter includes at least one of the target deep conversion rate, the actual deep conversion rate, and the estimated deep conversion rate; the deep conversion behavior includes at least one of the next-day retention, three-day retention, and seven-day retention.

[0044] In a possible implementation manner, the estimated resource acquisition amount module includes:

[0045] A shallow target resource acquisition amount unit, configured to calculate the shallow target resource acquisition amount according to the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate of the recommendation information;

[0046] A floating ratio calculation unit, configured to determine the floating ratio of the shallow target resource acquisition amount when the recommendation information achievement rate is within a preset range according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate;

[0047] An estimated resource acquisition amount calculation unit, configured to calculate the estimated resource acquisition amount of the recommendation information according to the shallow target resource acquisition amount and the floating ratio.

[0048] In a possible implementation manner, the floating ratio calculation unit includes:

[0049] The first coefficient calculation sub-unit is used to obtain the first quotient of the first ratio and the first calibration coefficient, and obtain the first coefficient according to the difference between the first quotient and the preset reference value;

[0050] The floating value calculation sub-unit is used to obtain the second quotient of the second ratio and the second calibration coefficient, determine the difference between the second quotient and the preset reference value, and obtain the floating value of the second coefficient in the current cycle compared with the second coefficient in the previous cycle according to the product of the difference and the preset step size;

[0051] The second coefficient calculation sub-unit obtains the second coefficient in the previous cycle and combines the floating value to obtain the second coefficient in the current cycle;

[0052] The coefficient product sub-unit is used to obtain the floating ratio according to the first coefficient and the second coefficient in the current cycle.

[0053] In a possible implementation, the floating ratio calculation unit further includes:

[0054] The relationship determination sub-unit is used to determine the magnitude relationship between the preset cost of the recommended recommended information and the initial preset cost of the recommender according to the magnitude relationship between the first coefficient and the preset threshold;

[0055] The cost increase notification unit is used to send a cost increase notification to the recommender corresponding to the recommended information if the preset cost of the recommended recommended information is higher than the initial preset cost of the recommender;

[0056] The floating value calculation sub-unit is specifically used to: if the cancellation recommendation instruction sent by the recommender according to the cost increase notification is not received, obtain the second quotient of the second ratio and the second calibration coefficient.

[0057] In a possible implementation, the first coefficient calculation sub-unit further includes a calibration coefficient determination sub-unit for obtaining the first calibration coefficient. Specifically, the calibration coefficient determination sub-unit includes:

[0058] The actual cost per action acquisition sub-unit is used to acquire the actual cost per action of the recommended information in the current cycle;

[0059] The deviation calculation sub-unit is used to obtain the first calibration coefficient according to the deviation between the actual cost per action of the recommended information in the current cycle and the target cost per action.

[0060] In a possible implementation, the request acquisition module includes an actual deep conversion rate acquisition sub-module for acquiring the actual deep conversion rate. The actual deep conversion rate acquisition sub-module includes:

[0061] The first parameter acquisition unit is used to acquire the estimated deep conversion rate of the recommended information in the preset time window as the first parameter;

[0062] A second parameter acquisition unit, configured to acquire the deep conversion number and the activation return ratio of the recommendation information in a preset time window, and obtain a second parameter according to the conversion number and the activation return ratio of the recommendation information in the preset time window;

[0063] A deep conversion rate calculation unit, configured to calculate an actual deep conversion rate according to the first parameter and the second parameter.

[0064] In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the method provided in the first aspect are implemented.

[0065] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method provided in the first aspect are implemented.

[0066] In a fifth aspect, an embodiment of the present invention provides a computer program, which includes computer instructions. The computer instructions are stored in a computer-readable storage medium. When a processor of a computer device reads the computer instructions from the computer-readable storage medium, the processor executes the computer instructions, so that the computer device executes the steps of the method provided in the first aspect.

[0067] The method, device, electronic device, and storage medium for delivering recommendation information provided by the embodiments of the present invention receive a recommendation information pull request from a terminal, obtain at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request, and determine the shallow conversion behavior parameter and the deep conversion behavior parameter of the recommendation information. Compared with the prior art that only estimates the resource acquisition amount based on the shallow conversion behavior parameter, by considering the deep conversion behavior, the delivery effect of the recommendation information can be evaluated more accurately, and the estimated resource acquisition amount of the recommendation information obtained thereby has the premise that the recommendation information achievement rate is within a preset range, which is in line with the long-term interests of the recommendation information alliance and the recommender. While the estimated resource acquisition amount of the recommendation information is more accurate, the probability of frequent price adjustment can be reduced, the recommendation information display resources can be saved, the delivery efficiency of the recommender can be improved, and the cost of delivering the recommendation information can be reduced. Description of the Drawings

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description in the embodiments of the present application.

[0069] Figure 1 A schematic diagram of a conversion process provided for an embodiment of the present application;

[0070] Figure 2The architecture diagram of a recommendation information delivery system provided by an embodiment of the present application;

[0071] Figure 3 The schematic diagram of the recommendation information delivery process of the recommendation information delivery system provided by an embodiment of the present application;

[0072] Figure 4 The schematic diagram of the process of a recommendation information delivery method provided by an embodiment of the present application;

[0073] Figure 5 The schematic diagram of the process of a recommendation information delivery method provided by another embodiment of the present application;

[0074] Figure 6 The schematic diagram of the process of a recommendation information delivery method provided by an embodiment of the present application;

[0075] Figure 7 The structural schematic diagram of a recommendation information delivery device provided by an embodiment of the present application;

[0076] Figure 8 The structural schematic diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0077] The embodiments of the present application are described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and should not be construed as a limitation of the present invention.

[0078] Those skilled in the art of the present technology can understand that unless specifically stated otherwise, the singular forms "a", "an" and "the" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of the present application means the presence of features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The phrase "and / or" used herein includes all or any unit and all combinations of one or more related listed items.

[0079] The recommendation information delivery method, device, electronic device and storage medium provided by the present application are intended to solve the above technical problems of the prior art.

[0080] It should be understood that the present application provides a method for delivering recommended information based on Artificial Intelligence (AI) technology. This method is applicable to the delivery of online recommended information on websites. For example, it can be used to deliver recommended information on search engines, in information flow products, on video websites, and on TVs. The AI technology attempts to provide intelligent marketing strategies for recommenders, and at the same time provide quotes for recommended information positions that are more accurate and have a relatively stable conversion rate. This can not only increase the profits of enterprises, but also improve the stickiness of recommenders to a certain extent.

[0081] It can be understood that AI uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, including theories, methods, technologies, and application systems for perceiving the environment, acquiring knowledge, and using knowledge to obtain the best results. In other words, AI is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. AI also studies the design principles and implementation methods of various intelligent machines to enable machines to have the functions of perception, reasoning, and decision-making. Among them, AI technology is an interdisciplinary subject with a wide range of fields, including both hardware-level technologies and software-level technologies. AI basic technologies generally include technologies such as sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. AI software technologies mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0082] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0083] To facilitate the understanding of the method provided by the embodiments of the present application, first, the nouns involved in the embodiments of the present application will be introduced:

[0084] A traffic media platform is a platform that can provide traffic and is also a carrier for displaying recommended information. It usually includes media, websites, and applications. Traffic media platforms can participate in the profit sharing of recommended information. For example, QQ, WeChat, and QQ Browser can all be used as traffic media platforms.

[0085] Recommenders (advertisers): refer to individuals, enterprises, or units that want to promote their brands or products through recommended information. For example, Zhang San, a citizen of Beijing, BMW, Intel, Mengniu, Tencent, etc.

[0086] Audience: the people who "consume" the recommended information, that is, consumers and users.

[0087] The Recommended Information Alliance Connect refers to the Network Recommended Information Alliance, which means aggregating small and medium-sized online media resources (also known as alliance members, such as small and medium-sized websites, personal websites, WAP sites, etc.) to form an alliance. Through the alliance, it helps the recommender to achieve the placement of recommended information and conducts monitoring and statistics on the placement data of recommended information. The recommender then pays the recommended information fee to the alliance members according to the actual effect of the network recommended information. This is a form of network recommended information organization placement.

[0088] eCPM (effective cost per mile), which is the display revenue of the media, generally refers to the recommended information revenue that the media can obtain for every one thousand displays. For example, if the number of displays of the recommended information is 45,000 and the revenue obtained is 180 yuan, then the revenue per thousand displays is equal to 180 / 45, that is, 4 yuan. Based on this parameter, the effect of placing recommended information can be analyzed, and the recommended information can be optimized and adjusted accordingly to increase revenue.

[0089] The estimated resource acquisition volume is the display revenue calculated by this application after optimizing the existing eCPM calculation method and using the improved eCPM calculation formula. Using this estimated resource acquisition volume to evaluate the revenue of media recommended information is more reasonable than the existing method of calculating the revenue of recommended information using eCPM, improves the utilization of traffic value, and has high commercial value.

[0090] CPA (Cost Per Action), the cost per action, where the action can be registration, interaction, download, order placement, purchase, etc.; CPA = total cost / conversion volume. For example, within a certain period, the cost of a recommender investing in the recommended information of a certain product is 6,000 US dollars, the number of exposures of this recommended information is 600,000, the number of clicks is 60,000, and the number of conversions (such as activation) is 1,200. Then the cost per action of this recommended information is: CPA = 6,000 / 1,200 = 5 US dollars. In the embodiments of this application, the total cost is also referred to as the resource occupancy volume, and the resources can be money, human resources, time, etc.

[0091] tCPA, the target cost per action, and this value is set by the recommender.

[0092] CTR (Click-Through-Rate), the click-through rate, which can also be simply referred to as the click rate, is the actual number of clicks of the recommended information (divided by the display volume of the recommended information (Show content)); CTR = click volume / display volume;

[0093] pCTR (Predict CTR), the predicted click rate, is the click rate predicted by the prediction model based on the characteristics of the recommended information itself, the target audience, and the historical click rate.

[0094] CVR (Conversion Rate), the conversion rate, which refers to the conversion rate from when a user clicks on a recommended message to becoming a validly activated or registered user or even a paying user; CVR = number of conversions / number of clicks, which is an indicator for measuring the effectiveness of CPA recommended messages;

[0095] pCVR (Predict CVR), the predicted conversion rate, which is estimated by a prediction model based on the characteristics of the recommended message, the target audience, and historical conversion rates.

[0096] DCVR (deep CVR), is a concept derived from the definition of the conversion rate in the embodiments of this application, the deep conversion rate, which can be the next-day retention rate, three-day retention rate, seven-day retention rate, etc. in this solution.

[0097] pDCVR (Predict DCVR): the predicted deep conversion rate, that is, the predicted deep conversion rate.

[0098] tDCVR (targetDCVR), the target deep conversion rate, that is, the target value of the deep conversion rate, which is set by the recommender.

[0099] DCVR now : the actual deep conversion rate;

[0100] cnt avtive : the accumulation of the conversion rate within a time window;

[0101] ratio active : the activation return ratio, which represents the ratio of the returned users to the activated users;

[0102] Resource consumption: that is, the customer acquisition cost of the recommender (for example, in the CPA model, getting 100 users for 1000 yuan, the cost is 10);

[0103] Number of active users: users who log in again within a period of time or log in multiple times in the next few days.

[0104] Price adjustment: when setting the preset cost of the recommended message, use a factor to control the actual bid ecpm. In the embodiments of this application, the core of price adjustment is to control the achievement rate.

[0105] Achievement Cost: If |Actual Cost of Recommended Information - Preset Cost (set by the recommender for this recommended information)| / Preset Cost ≤ Specified Threshold α (α < 1), it is called achievement cost. For a recommended information, if the ratio of the difference between the actual cost and the preset cost of the recommended information to the preset cost is less than the preset threshold, then it can be considered that this recommended information meets the achievement cost. For example, if the actual cost of a certain recommended information is 20 and the preset cost is 15, and the specified threshold is 30%, then through the above formula, it can be known that this recommended information meets the achievement cost. The preset cost can be understood as the preset cost set by the recommender for the recommended information.

[0106] The achievement rate can be calculated in two ways: based on resource consumption and based on quantity. For the calculation based on resource consumption, the formula is: Resource Consumption of Recommended Information with Achievement Cost / Resource Consumption of All Recommended Information * 100%. For the calculation based on quantity, the formula is: Number of Recommended Information with Achievement Cost / Number of All Recommended Information * 100%.

[0107] It should be understood that the actual cost of recommended information usually refers to the cost per thousand times. Therefore, the resource consumption of recommended information can be calculated by multiplying the actual cost of the recommended information by the conversion times of the recommended information.

[0108] For example, there are 3 recommended information with achievement, the resource consumption of these 3 recommended information is 300,000 yuan, and at the same time, the total number of recommended information is 5, and the resource consumption of these 5 recommended information is 1,000,000 yuan. Then the achievement rate calculated based on resource consumption is 30%, and the achievement rate calculated based on the number of recommended information is 60%.

[0109] The conversion preset cost is the core of information flow recommended information and one of the factors that have the greatest impact on the effect of recommended information. There are two common preset cost methods for network recommended information: click preset cost and conversion preset cost. The click preset cost is "how much money the recommender is willing to pay at most for each click", which is good for controlling the click cost; the conversion preset cost is "how much money the recommender is willing to pay at most for each conversion", which is good for controlling the conversion cost.

[0110] For example: The recommender wants the activation cost to be 5 yuan. Then the recommender needs to control the click preset cost to control the activation cost. There is also the conversion rate from click to activation involved, so the formula between them is:

[0111] Click Cost ÷ Conversion Rate from Click to Activation = Activation Cost

[0112] Although the cost per click is easy to control, the click-to-activation rate can only be estimated, so the activation cost is often out of control. For example, a recommended message is planned to cost 0.3 yuan per click, and the activation cost is 5 yuan. However, what often happens is that the preset cost per click remains 0.3 yuan, but the activation cost becomes 10 yuan, doubling the cost.

[0113] The conversion preset cost can solve the above problem, because the media was originally only responsible for the click cost to meet the recommender's preset cost requirements, but now the conversion cost must meet the recommender's preset cost. What is "conversion" is customizable by the recommender. Activation, registration, ordering, payment... can all be considered conversions. However, the conversion preset cost brings new problems: the cost and the preset cost often do not match, especially the cost is often higher than the preset cost. For ease of understanding, please refer to Figure 1 , is a schematic diagram of a conversion process provided in an embodiment of the present application, such as Figure 1 As shown,

[0114] QQ has many users, they have different characteristics, men and women, young and old, with different interests and hobbies, and the shapes are used to represent people with different characteristics. Some users are triangles, some are five-pointed stars, some are squares, and some are circles.

[0115] The process of placing recommendation information is divided into two stages. In the first stage (step ① in the figure), first find a group of people, show them the recommendation information, and see their reactions. Some of them click, some do not, and some convert after clicking. The media needs to focus on those who convert. Here, circles and hearts are used to represent people who convert.

[0116] After finding the converted people, we need to further analyze their characteristics, which leads to the second stage (step ② in the figure). The media speculates that those who have converted should like this recommendation information more, so will other circular and heart-shaped users also like this recommendation information? By showing this recommendation information to more circular and heart-shaped users, the system forms a model for finding people. This is the principle of machine learning. First find some converted people, then analyze their characteristics, build a model, and then find people similar to them based on the model.

[0117] There are three reasons why the display revenue of existing recommendation information delivery process is unstable:

[0118] Reason 1: Generalized 2-price deduction (GSP nd The second-price deduction refers to charging the recommended information with the second highest preset cost to the recommended information that won the request.

[0119] Ignoring other factors that affect costs, GSP will result in the actual display revenue of recommended information not being higher than the preset costs of the recommenders.

[0120] Reason 2: The deviation in pCVR prediction leads to the failure to achieve the display revenue

[0121] The deviation in pCVR prediction refers to the deviation between the pCVR of the optimization target of recommended information and the conversion rate. Ignoring other factors that affect costs, the deviation in pCVR prediction will result in the actual revenue of recommended information (which can be characterized by the estimated resource acquisition volume) not being equal to the preset costs of the recommenders.

[0122] Reason 3: Conversion reflux leads to fluctuations in display revenue

[0123] The long conversion reflux time affects the real-time cost fluctuations of recommended information.

[0124] The current calculation scheme for display revenue generally only considers shallow conversion behaviors. Shallow conversion behaviors generally include more forward behaviors such as clicks and activations. The preset costs for shallow goals, that is, the preset costs when achieving the goals set based on shallow behaviors, are generally measured by activation costs. In related technologies, the calculation formula for the display costs (shallow goal resource acquisition volume) of the media is as follows:

[0125] eCPM = λ × tCPA × pCTR × pCVR

[0126] Among them, λ is a preset parameter, which can be the deviation between the actual CPA and the target CPA of shallow conversion behaviors. The actual CPA and the target CPA of shallow conversion behaviors, that is, in the case of achieving shallow goals (such as clicks), the actual cost per action and the target cost per action.

[0127] Currently, the method based on the shallow goal resource acquisition volume easily leads to sudden changes in the achievement rate, and then causes the recommenders to frequently and significantly adjust prices, affecting efficiency and experience.

[0128] The following uses specific embodiments to elaborate in detail on the technical solutions of this application and how the technical solutions of this application solve the above technical problems. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The following will describe the embodiments of this application in conjunction with the drawings.

[0129] Figure 2 The architecture diagram of a recommended information delivery system provided by an embodiment of this application is as Figure 2As shown in the figure, the recommendation information alliance 11 is an alliance formed by aggregating multiple traffic media platforms. In the figure, the recommendation information alliance has one traffic media platform. It should be noted that the number of traffic media platforms in the embodiments of the present application is not specifically limited. Here, for the convenience of description, it is set to 3 traffic media platforms 111. The background server of one of the traffic media platforms is the traffic media platform server 112. When the terminal 21 accesses the traffic media platform server 112, the traffic media platform server 112 provides traffic services. It can be understood that the traffic services can be to display web pages to the terminal. The traffic media platform server is connected to the recommendation information delivery server 31 through a network. The recommendation information delivery server 31 is used to execute the recommendation information delivery methods in the following embodiments and send the finally determined recommendation information to the traffic media platform server 112. The traffic media platform server 112 returns the web page with the recommendation information to the terminal.

[0130] Both the traffic media platform server and the recommendation information delivery server can be independent physical servers, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0131] The execution method of the server in the embodiments of the present application can be completed in the form of cloud computing. Cloud computing is a computing model that distributes computing tasks on a resource pool composed of a large number of computing devices, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides resources is called the "cloud". The resources in the "cloud" seem to be infinitely expandable to users, and can be obtained at any time, used on demand, expanded at any time, and paid according to usage.

[0132] As a basic capability provider of cloud computing, a cloud computing resource pool (abbreviated as a cloud platform, generally referred to as an IaaS (Infrastructure as a Service) platform) will be established, and various types of virtual resources will be deployed in the resource pool for external customers to select and use. The cloud computing resource pool mainly includes: computing devices (virtual machines containing operating systems), storage devices, and network devices.

[0133] According to the logical function division, on the IaaS (Infrastructure as a Service) layer, the PaaS (Platform as a Service) layer can be deployed, and on top of the PaaS layer, the SaaS (Software as a Service) layer can be deployed. Alternatively, the SaaS can be directly deployed on the IaaS. PaaS is the platform for software operation, such as databases, web containers, etc. SaaS is various business software, such as web portals, SMS mass senders, etc. Generally speaking, SaaS and PaaS are the upper layers relative to IaaS.

[0134] Figure 3 The following is a schematic diagram of the delivery process of the recommended information for the recommended information delivery system provided by the embodiments of this application, as Figure 3 shown, including:

[0135] S1. Multiple recommenders 41 send a delivery request for the traffic media platform 111 and the materials of the recommended information to be delivered to the recommended information alliance 11. The delivery request includes the initial preset cost of the recommender.

[0136] S2. The terminal 21 sends a recommended information pulling request to the traffic media platform server 112. The recommended information pulling request may include user characteristics, such as the user's age, gender, the unique identifier of the terminal, etc., so as to enable the traffic media platform server 112 or the recommended information delivery server 31 to estimate some parameters based on the user characteristics.

[0137] S3. The traffic media platform server 112 forwards the recommended information pulling request to the recommended information delivery server 31;

[0138] S4. The recommended information delivery server 31 obtains at least one recommended information corresponding to the recommended information pulling request from the recommended information alliance 11;

[0139] S5. For each of the at least one recommended information corresponding to the recommended information pulling request, the recommended information delivery server 31 obtains the estimated resource acquisition amount of the recommended information when the recommended information achievement rate of the recommended information meets the preset conditions, and determines the target recommended information based on the estimated resource acquisition amounts of the at least one recommended information. The recommended information delivery server sends the information of the target recommended information to the recommended information alliance;

[0140] S6. The recommended information alliance distributes the materials of the target recommended information to the traffic media platform server;

[0141] S7. The traffic media platform server loads the target recommendation information into the web page, obtains the web page data including the target recommendation information, and returns the web page data to the terminal 21. The terminal displays the web page including the target recommendation information.

[0142] Optionally, the recommendation information delivery server in the above embodiment can also be used as a module in the recommendation information alliance 11.

[0143] Figure 4 The flowchart of a method for delivering recommendation information provided by an embodiment of the present application is shown in Figure 4 as follows, including:

[0144] S101. Receive a recommendation information pull request from the terminal, and obtain at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request, where the recommendation information pull request is a request for the terminal to obtain the recommendation information displayed in the web page.

[0145] Among them, the terminal in the embodiment of the present application can be various different electronic devices, such as mobile phones, tablets, computers, wearable electronic devices, etc. In practical applications, a web page can include at least one recommendation information display page. When there are multiple recommendation information display pages, the embodiment of the present invention can execute the Figure 3 method shown respectively for each recommendation information display page to determine the estimated resource acquisition amount of each recommender for each recommendation information display area.

[0146] Among them, the web page content in the web page can be composed of multiple display layers, and the recommendation information display page can be located in any display layer. For example: The first web page is composed of a first display layer and a second display layer, and the first display layer is above the second display layer. Then the recommendation information display page can be located in the first display layer or in the second display layer. The present invention does not make a limitation here.

[0147] Specifically, Figure 4 the method shown in Figure 2 can be applied to the recommendation information delivery server shown above. The recommendation information delivery server can also be communicatively connected to the terminal. The above access request can be sent by the terminal to the recommendation information delivery server. After receiving the access request, the recommendation information delivery server obtains the above recommendation information pull request from the recommendation information alliance.

[0148] Among them, the recommendation information display area can have various different attribute information, such as: size, display method, relative position in the web page, etc.

[0149] S102. For each of at least one piece of recommendation information, determine the shallow conversion behavior parameter and the deep conversion behavior parameter of each piece of recommendation information, and obtain the estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter.

[0150] The shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior. The shallow behavior conversion parameter includes at least one of the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate; the shallow conversion behavior includes at least one of exposure, click, and activation.

[0151] The deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the deep conversion behavior parameter includes at least one of the target deep conversion rate, the actual deep conversion rate, and the estimated deep conversion rate; the deep conversion behavior includes at least one of the next-day retention, the three-day retention, and the seven-day retention.

[0152] It should be understood that exposure is the behavior of a visitor seeing the recommendation information. Click is the behavior of a visitor clicking on the recommendation information after seeing it. Activation is the behavior of a user downloading, registering as a member, etc. after clicking on the recommendation information. It can be seen that since exposure, activation, and download are all instantaneous behaviors, the interaction degree between the user and the recommendation information is relatively low. And the next-day retention is the ratio of users who stay among the newly added users compared to the previous day. Similarly, the meanings of the three-day retention and the seven-day retention can be known. Obviously, the next-day process, the three-day retention, and the seven-day retention are behaviors that occur over a relatively long period of time, and such behaviors mean a relatively high interaction degree between the user and the recommendation information.

[0153] From the above explanations of each technical term, it can be seen that the target cost per action and the target deep conversion rate of the recommendation information are set by the recommender, while the estimated click-through rate pCTR, the estimated shallow conversion rate pCVR, and the estimated deep conversion rate pDCVR are estimated through corresponding prediction models.

[0154] It should be understood that after defining the conversion actions corresponding to the shallow target and the deep target, the corresponding prediction can be made by statistically analyzing the historical data of the recommendation information. The following takes obtaining the estimated click-through rate pCTR as an example for explanation.

[0155] S200. Obtain at least one of the following features:

[0156] The recommendation information feature of the recommendation information uploaded by the recommender, the display environment feature of the recommendation information display page, the feature of the terminal, and the user feature of the user of the terminal.

[0157] Among them, the recommended information features may include: the duration of the recommended information, the popularity of the people in the recommended information, the quality score of the recommended information, etc. These recommended information features will all affect whether the people who view the recommended information will click on the recommended information. Generally speaking, when the duration of the recommended information is appropriate, the popularity of the people in the recommended information is higher, and the quality score of the recommended information is higher, the probability that the recommended information is clicked is greater.

[0158] The display environment features may include: the types and industries of other web page contents except the recommended information in the web page where the recommended information is located, etc. Among them, the types of web page contents may include: text, audio, still pictures, dynamic pictures, videos, etc. The industries to which the web page contents belong may include: automobiles, beauty, entertainment, film and television, etc. It can be understood that when the recommended information matches the display environment features, the recommended information can obtain a higher click-through rate. For example: when a certain automobile recommended information is on a web page carrying a large amount of automobile industry text and still pictures, the automobile recommended information can obtain a higher click-through rate.

[0159] The features of the terminal may include: the type of the terminal's operating system, the name of the browser for loading the web page, etc.

[0160] The user features of the user of the terminal may include: the age, gender, education level, monthly income, etc. of the user.

[0161] It can be understood that the click probabilities of different operating system types, browsers, and different users for recommended information in different industries are also different. Therefore, the click-through rate of the recommended information can be further accurately predicted according to the features of the terminal and the user features of the user of the terminal.

[0162] S210. Input the obtained features into the click-through rate prediction model to obtain the predicted click-through rate pCTR output by the click-through rate prediction model, and determine this predicted click-through rate as the predicted click-through rate of the recommended information uploaded by this recommender on the recommended information display page.

[0163] Optionally, the above click-through rate prediction model may include: an offline model and / or an online model.

[0164] The click-through rate prediction model can be obtained by means of machine learning on a large amount of training data. In practical applications, some recommended information can be launched for testing, and users can be allowed to browse and click to collect the recommended information data as training data. Specifically, the embodiments of the present invention can control the testing process of the recommended information launched online in the following manner:

[0165] A) The number of display exposures of a single recommended information is not higher than a preset number of displays;

[0166] B) Stop testing 24 hours after the recommended information is launched online;

[0167] C) The number of times the same recommended information is displayed and exposed to the same user is not more than 3 times.

[0168] In the embodiment of the present application, when calculating the estimated resource acquisition amount of the recommended information, on the one hand, the conversion rate is divided into a shallow conversion rate and a deep conversion rate, so as to more accurately evaluate the conversion behavior. On the other hand, since the calculation period of the deep conversion rate is longer than that of the shallow conversion rate, the finally calculated estimated resource acquisition amount of the recommended information can also be maintained for a longer time, avoiding frequent price adjustments by the recommender. On the other hand, the estimated resource acquisition amount of the recommended information obtained in the embodiment of the present application needs to meet the condition that the recommended information achievement rate is within a preset range. Specifically, the recommended information achievement rate is a parameter that both the recommended information alliance and the recommender need to pay attention to. If the recommended information achievement rate is low for a long time, the willingness of the recommender to continue publishing the recommended information will decrease, ultimately leading to the loss of the recommender. If the recommended information achievement rate exceeds the recommender's target for a long time, it means that the interests of the recommended information alliance are damaged, which will affect the pricing power of the recommended information alliance for the recommended information in the long run. Based on this, the embodiment of the present application sets a prerequisite for the estimated resource acquisition amount of the recommended information - the recommended information achievement rate is within a preset range, so as to balance the interests of the recommended information alliance and the recommender and lay a foundation for the development of the recommended information alliance.

[0169] Optionally, the recommended information achievement rate in the embodiment of the present application can be set in the range of 90% - 95%. As long as the calculated recommended information achievement rate is within this range, it is considered that the recommended information achievement rate meets the preset conditions. By setting the range, it is also possible to buffer the fluctuations in the estimated resource acquisition amount of the recommended information, further improving the stability of the estimated resource acquisition amount of the recommended information.

[0170] In practical applications, steps S101 and S102 can be completed within a short time, so as not to delay the display of the web page on the terminal too much.

[0171] S103. Determine the target recommended information from at least one recommended information according to the estimated resource acquisition amounts of multiple recommended information, and return the data corresponding to the target recommended information to the terminal, so that the terminal displays the target recommended information.

[0172] Specifically, after determining the estimated resource acquisition amounts of each recommender, the estimated resource acquisition amounts can be further sorted from high to low, and the recommender with the highest estimated resource acquisition amount is used as the target recommender. The recommended information of the target recommender is also called the target recommended information. The target recommended information is set on the recommended information display page, and the web page data containing the target recommended information is returned to the terminal, so that the terminal displays the web page containing the recommended information, realizing the delivery of the recommended information.

[0173] The method for delivering recommended information provided by the embodiments of the present application receives a request for pulling recommended information from a terminal, obtains at least one piece of recommended information corresponding to the request for pulling recommended information, and determines the shallow conversion behavior parameters and deep conversion behavior parameters of the recommended information. Compared with the prior art that only calculates the estimated resource acquisition volume based on the shallow conversion behavior parameters, by considering the deep conversion behavior, it can more accurately evaluate the delivery effect of the recommended information. Moreover, the estimated resource acquisition volume of the recommended information obtained thereby has the premise that the achievement rate of the recommended information is within a preset range, which is in line with the long-term interests of the recommended information alliance and the recommender. While the estimated resource acquisition volume of the recommended information is more accurate, it can also reduce the occurrence probability of frequent price adjustments, save the display resources of the recommended information, improve the delivery efficiency of the recommender, and reduce the cost of delivering the recommended information.

[0174] Figure 5 is a schematic flowchart of a method for delivering recommended information provided by another embodiment of the present application, as Figure 5 shown, including:

[0175] S201. Receive a request for pulling recommended information from a terminal, and obtain at least two pieces of recommended information corresponding to the request for pulling recommended information.

[0176] This step is the same as or similar to the content of step S101 in Figure 4 the embodiment, and will not be elaborated here.

[0177] S202. Obtain the estimated resource acquisition volume of the recommended information when the achievement rate of the recommended information meets the preset conditions, including S2021 to S2023. Specifically:

[0178] S2021. Calculate the shallow target resource acquisition volume according to the target cost per action tCPA, the estimated click-through rate pCTR, and the estimated shallow conversion rate pCVR of the recommended information;

[0179] Specifically, the embodiments of the present application can refer to the existing eCPM calculation method, and take the product of the target cost per action tCPA, the estimated click-through rate pCTR, and the estimated shallow conversion rate pCVR as the shallow target resource acquisition volume.

[0180] S2022. Determine the floating ratio of the shallow target resource acquisition volume when the achievement rate of the recommended information meets the preset conditions according to the first ratio of the estimated deep conversion rate pDCVR to the target deep conversion rate tDCVR, and the second ratio of the actual deep conversion rate DCVR now to the target deep conversion rate tDCVR.

[0181] In the process of calculating the floating ratio of the acquisition volume of shallow target resources in the embodiments of the present application, the ratios of the estimated deep conversion rate to the target deep conversion rate and the actual deep conversion rate to the target deep conversion rate are considered respectively. The reason for considering the above two ratios is that the estimated deep conversion rate reflects the future change trend of the deep conversion rate. If the estimated deep conversion rate is lower than the target deep conversion rate, it means that it is difficult to meet the expectations of the recommender under the current placement strategy of the recommended information, and the placement strategy of the recommended information needs to be adjusted, and reducing the price is beneficial to retaining the recommender. If the estimated deep conversion rate is higher than the target deep conversion rate, it means that it is very likely to meet the expectations of the recommender under the current placement strategy of the recommended information. At this time, if the placement strategy of the recommended information is not adjusted, it is very likely that the actual conversion rate will be higher than the target of the recommender, and then the actual cost per action will be lower than the estimated resource acquisition volume of the recommender. This result will affect the revenue of the recommended information alliance. Therefore, the price can be appropriately increased to maintain the achievement rate of the recommender's recommended information at a normal level, which is more conducive to maintaining customers.

[0182] Further, if the actual deep conversion rate is lower than the target deep conversion rate, it means that under the current placement strategy of the recommended information, the actual deep conversion rate still cannot meet the expectations of the customer. However, this may be because the placement has just started, the target population of the recommended information placement is not clear, or the relevant information of the deep conversion rate has not been counted in time. Therefore, it is necessary to make a judgment by combining the magnitudes of the estimated deep conversion rate and the target deep conversion rate.

[0183]

[0184]

[0185] Table 1 is the preset cost adjustment logic table of the embodiments of the present application

[0186] Table 1 is the preset cost adjustment logic table of the embodiments of the present application. Next, the preset cost adjustment strategy of the embodiments of the present application will be specifically described with respect to Table 1:

[0187] When the actual deep conversion rate is greater than the target deep conversion rate and the estimated deep conversion rate is greater than the target deep conversion rate, since the estimated deep conversion rate is greater than the target deep conversion rate, it indicates that the final deep conversion rate is likely to be higher than the recommender's expectation. Therefore, the preliminary preset cost adjustment strategy is determined to be a price increase (i.e., increasing the preset cost). Both price increases and price cuts (i.e., reducing the preset cost) are relative to the recommender's initial preset cost. It should be understood that when the recommender launches the recommended information, an initial preset cost will be set to facilitate adjustments based on this initial preset cost. At the same time, since the actual deep conversion rate is also greater than the target deep conversion rate, it means that the current trend of the conversion rate change is spreading. Therefore, a relatively large price increase is determined. If the placement strategy of the recommended information is not adjusted at this time, it will inevitably result in a significantly low cost per action of the recommended information. Therefore, a relatively large price increase will not cause rejection from the recommender and is also conducive to maintaining the achievement rate of the recommender's recommended information at the preset level.

[0188] When the actual deep conversion rate is less than the target deep conversion rate and the estimated deep conversion rate is greater than the target deep conversion rate, first, according to the above logic, the preliminary adjustment strategy is still a price increase. However, since the actual deep conversion rate is less than the target deep conversion rate, it means that the current trend of the conversion rate change is converging. Therefore, a relatively small price increase is determined.

[0189] When the actual deep conversion rate is greater than the target deep conversion rate and the estimated deep conversion rate is less than the target deep conversion rate, since the estimated deep conversion rate is less than the target deep conversion rate, the preliminary preset cost adjustment strategy is a price cut. At the same time, since the actual deep conversion rate is greater than the target deep conversion rate, it means that the current trend of the conversion rate change is spreading, and a relatively large price cut is required to maintain the recommender.

[0190] When the actual deep conversion rate is less than the target deep conversion rate and the estimated deep conversion rate is less than the target deep conversion rate, since the estimated deep conversion rate is less than the target deep conversion rate, the preliminary preset cost adjustment strategy is also a price cut. At the same time, since the actual deep conversion rate is greater than the target deep conversion rate, it means that the current trend of the conversion rate change is converging, and only a relatively small price cut is required.

[0191] When the actual deep conversion rate is equal to the target deep conversion rate, it indicates that the trend of the conversion rate change is stable. At this time, it can be determined whether to increase or decrease the price based on the comparison between the estimated deep conversion rate and the target deep conversion rate. However, the specific amplitude can be calculated through the subsequent embodiments.

[0192] When the estimated deep conversion rate is equal to the target deep conversion rate, it indicates that the placement strategy of the recommended information at this time is consistent with the recommender's expectation, and there is no need to adjust the preset cost.

[0193] S2023. Calculate the estimated resource acquisition volume of the recommended information based on the acquisition volume of the shallow target resources and the floating ratio.

[0194] Specifically, in the embodiments of the present application, the product of the acquisition volume of the shallow target resources and the floating ratio can be used as the estimated resource acquisition volume of the recommended information. For example, if the acquisition volume of the shallow target resources is a and the floating ratio is 20%, it means that on the basis of the acquisition volume of the shallow target resources, it floats upward by 20%. Then the estimated resource acquisition volume of the recommended information is 1.2a. Also, for example, when the floating ratio is -10%, it means that on the basis of the acquisition volume of the shallow target resources, it floats downward by 10%, that is, the estimated resource acquisition volume of the recommended information is 0.9a.

[0195] S203. Determine the target recommended information from at least one recommended information according to the estimated resource acquisition volumes of multiple recommended information, and return the data of the target recommended information to the terminal so that the terminal can display the target recommended information.

[0196] This step is the same as or similar to Figure 4 the content of step S103 in the embodiment, and will not be elaborated here.

[0197] Based on the above embodiments, as an alternative embodiment, according to the first ratio of the estimated deep conversion rate pDCVR to the target deep conversion rate tDCVR, and the second ratio of the actual deep conversion rate DCVR now to the target deep conversion rate tDCVR, determine the floating ratio of the acquisition volume of the shallow target resources when the recommendation achievement rate of the recommended information meets the preset conditions, including:

[0198] S301. Obtain the first quotient value of the first ratio and the first calibration coefficient, and obtain the first coefficient according to the difference between the first quotient value and the preset reference value.

[0199] To better measure the influence of the estimated deep conversion rate on the floating ratio, the present application also sets a first calibration coefficient for the first ratio. The first calibration coefficient is used to characterize the deviation between the actual cost per action CPA and the target cost per action CPA.

[0200] Step S301 can be expressed by the following formula:

[0201]

[0202] Among them, W1 represents the first coefficient, λ1 represents the first calibration coefficient, and R represents the preset reference value, which can be 1. That is to say, the first coefficient actually measures the difference of the first ratio compared with the case where the estimated deep conversion rate is the same as the target deep conversion rate after considering the deviation between the actual cost per action and the target cost per action.

[0203] S302. Obtain the second quotient value of the second ratio and the second calibration coefficient, determine the difference between the second quotient value and the preset reference value, and obtain the floating value of the second coefficient in the current period compared with the second coefficient in the previous period according to the product of the difference and the preset step size.

[0204] Step S302 can be expressed by the following formula:

[0205]

[0206] Where S represents the floating value of the second coefficient in the current period compared with the second coefficient in the previous period, λ2 represents the second calibration coefficient. Optionally, the second calibration coefficient in the embodiments of the present application can be 1.2, and step represents the step size. The embodiments of the present application do not specifically limit the duration of one period, which can be half a day, one day, or even one week.

[0207] S303. Obtain the second coefficient in the previous period, and obtain the second coefficient in the current period by combining the floating value; for example, if the second coefficient in the previous period is defined as W2', then the second coefficient W2 in the current period = W2' + S.

[0208] S304. Obtain the floating ratio according to the first coefficient and the second coefficient in the current period.

[0209] Optionally, the embodiments of the present application can use the product of the first coefficient and the second coefficient in the current period as the floating ratio.

[0210] In the embodiments of the present application, by obtaining the first quotient value of the first ratio and the first calibration coefficient, obtaining the first coefficient according to the difference between the first quotient value and the preset reference value, obtaining the second quotient value of the second ratio and the second calibration coefficient, determining the difference between the second quotient value and the preset reference value, and obtaining the floating value of the second coefficient in the current period compared with the second coefficient in the previous period according to the product of the difference and the preset step size, obtaining the second coefficient in the previous period, obtaining the second coefficient in the current period by combining the floating value, obtaining the floating ratio of the acquisition amount of the shallow target resources after considering the deep target according to the first coefficient and the second coefficient in the current period, and finally combining the acquisition amount of the shallow target resources to obtain a more accurate estimated resource acquisition amount of the recommended information.

[0211] As can be seen from Table 1, the first ratio can be used to initially judge whether the preset cost will increase or decrease. Therefore, as an optional embodiment, before obtaining the second quotient value of the second ratio and the second calibration coefficient, it further includes:

[0212] Determine the size relationship between the recommended preset cost and the initial preset cost of the recommender according to the size relationship between the first coefficient and the preset threshold.

[0213] As can be seen from Table 1, when the estimated deep conversion rate is greater than the target deep conversion rate, a price increase is required; when the estimated deep conversion rate is less than the target deep conversion rate, a price cut is required; and when the estimated deep conversion rate is equal to the target deep conversion rate, the preset cost remains unchanged.

[0214] If the recommended preset cost is higher than the initial preset cost of the recommender, a cost increase notice is sent to the recommender corresponding to the recommendation information.

[0215] Since the embodiments of the present application can quickly determine the magnitude relationship between the recommended preset cost and the initial preset cost of the recommender after introducing the deep conversion behavior parameter, a cost increase notice is sent to the recommender when the recommended preset cost is higher than the initial preset cost, so as to determine the intention of the recommender. If the recommender is willing to increase the price, the recommender will provide a new quotation; if the recommender is not willing to increase the price, the notice can be ignored.

[0216] In addition, when the recommended preset cost is lower than the initial preset cost, a price cut notice can also be sent to the recommender to inform the recommender that they will compete again at a lower price.

[0217] Based on the above embodiments, obtaining the second product result of the second ratio and the second calibration coefficient includes:

[0218] If a cancellation recommendation instruction sent by the recommender according to the cost increase notice is not received, the second quotient value of the second ratio and the second calibration coefficient is obtained.

[0219] That is to say, after sending the cost increase notice to the recommender, the recommender can choose to send a cancellation recommendation instruction to indicate that they no longer compete for the recommendation information position. Therefore, if the cancellation recommendation instruction is not received, it is defaulted that the recommender still participates in the bidding, and step S302 will continue to be executed.

[0220] Figure 6 It is a schematic flowchart of a method for placing recommendation information provided by an embodiment of the present application. As Figure 6 shown, the process includes the following steps:

[0221] A1. Multiple recommenders 41 send a placement request for the traffic media platform and the materials of the recommendation information to be placed to the recommendation information alliance 11. The placement request is a placement request for the recommender's recommendation information on the recommendation information display page, and the placement request includes the initial preset cost of the recommender.

[0222] A2. The terminal 21 sends a recommendation information pull request to the traffic media platform server 112;

[0223] A3. The traffic media platform server 112 forwards the recommendation information pull request to the recommendation information placement server 31;

[0224] A4. The recommendation information delivery server 31 obtains at least one piece of recommendation information corresponding to the recommendation information pull request from the recommendation information alliance 11.

[0225] A5. For each piece of recommendation information among the at least one piece of recommendation information corresponding to the recommendation information pull request, the recommendation information delivery server 31 determines the magnitude relationship between the preset cost of the recommendation and the initial preset cost. If the preset cost to be recommended is higher than the initial preset cost of the recommender, a cost increase notice is sent to the recommender 41 corresponding to the recommendation information.

[0226] A6. If the recommender 41 sends a cancellation recommendation instruction to the recommendation information delivery server, after receiving the cancellation recommendation instruction, the recommendation information server excludes the recommender from the bidding for the recommendation information display page. If the recommender does not send a cancellation recommendation instruction to the recommendation information delivery server, the recommendation information delivery server continues to determine the estimated resource acquisition amount of the recommendation information and finally determines the target recommendation information from the estimated resource acquisition amounts of each piece of recommendation information.

[0227] A7. The recommendation information alliance 11 sends the material of the target recommendation information to the traffic media platform server 112.

[0228] A8. The traffic media platform server 112 loads the target recommendation information into the web page, obtains the web page data including the target recommendation information, and returns the web page data to the terminal 21. The terminal 21 displays the web page including the target recommendation information.

[0229] Based on the above embodiments, as an optional embodiment, the method for obtaining the first calibration coefficient includes:

[0230] S401. Obtain the actual cost per action tCPA of the recommendation information in the current period now ;

[0231] From the definition of the actual cost per action, that is, the quotient of the total actual cost and the actual conversion volume. Both of the above two parameters can be statistically obtained by the recommendation information alliance. It should be understood that the conversion behavior corresponding to the actual conversion volume used when calculating the first calibration coefficient can be a conversion behavior for shallow targets, such as clicks, activations, etc.

[0232] S402. Obtain the first calibration coefficient according to the deviation between the actual cost per action of the recommendation information in the current period and the target cost per action.

[0233] Based on the above embodiments, as an optional embodiment, the embodiments of the present application further provide a method for obtaining the actual deep conversion rate, including:

[0234] S501. Obtain the predicted deep conversion rate of the recommended information within a preset time window as the first parameter;

[0235] It should be understood that in the embodiments of the present application, a prediction model for predicting the deep conversion rate can be pre-constructed. The training principle of the prediction model for predicting the deep conversion rate is similar to the training principle of the model for predicting the click-through rate described above, including obtaining training samples and training labels. Each training sample is the characteristics of a user who has a deep conversion behavior for the recommended information and the characteristics of the recommended information, and the corresponding training label is the result information used to represent that the user has had a deep conversion behavior. Or the training sample can also be the characteristics of a user who has not had a deep conversion behavior for the recommended information and the characteristics of the recommended information, and the corresponding training label is the result information used to represent that the user has not had a deep conversion behavior.

[0236] Thus, when predicting the deep conversion rate, obtain the characteristics of the recommended information and the characteristics of the target population within the preset time window, and input the characteristics of the recommended information and the characteristics of each person in the population into the deep conversion rate prediction model, then the predicted deep conversion rate output by the deep conversion rate prediction model can be obtained. The embodiments of the present application do not limit the specific duration of the preset time window. For example, it can be half a day, one day, or even one week.

[0237] Further, since a recommended information will be displayed on multiple traffic channels simultaneously (for example, a mobile phone recommended information can be displayed on traffic channels such as QQ, QQ Music, QQ Browser, and WeChat at the same time), the prediction of the deep conversion rate can be specifically carried out for each traffic channel. Then, the first parameter in the embodiments of the present application can be represented by the following formula:

[0238]

[0239] where pDCVR i represents the predicted deep conversion rate of the recommended information on traffic channel i, and I represents the total number of traffic channels.

[0240] S502. Obtain the number of deep conversions and the activation return ratio of the recommended information within the preset time window, and obtain the second parameter according to the number of conversions and the activation return ratio of the recommended information in the current period.

[0241] The number of deep conversions is the number of conversion behaviors corresponding to the completion of the deep target, and can be at least one of the next-day retention, three-day retention, and seven-day retention. The activation return ratio refers to the ratio of the number of people who return after activation to the total number of activated people within a certain period of time.

[0242] The second parameter in the embodiments of the present application can be represented by the following formula:

[0243]

[0244] Among them, cnt active-i represents the number of deep conversions of the recommended information on traffic i, and ratio active-i represents the activation return ratio of the recommended information on traffic i.

[0245] S503. Calculate the actual deep conversion rate according to the first parameter and the second parameter.

[0246] Specifically, the quotient of the first parameter and the second parameter can be used as the actual deep conversion rate, and the actual deep conversion rate can be expressed by the following formula:

[0247] DCVR now = conv1 / conv2

[0248] The embodiments of the present application provide a device for delivering recommended information. As Figure 7 shown, the device may include: a request acquisition module 101, an estimated resource acquisition amount module 102, and an information return module 103. Specifically:

[0249] The request acquisition module 101 is configured to receive a recommended information pull request from a terminal, and obtain at least one recommended information corresponding to the recommended information pull request according to the recommended information pull request;

[0250] The estimated resource acquisition amount module 102 is configured to determine, for each of the at least one recommended information, the shallow conversion behavior parameter and the deep conversion behavior parameter of each recommended information, and obtain the estimated resource acquisition amount when the recommended information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter;

[0251] The information return module 103 is configured to determine a target recommended information from the at least one recommended information according to the estimated resource acquisition amounts of the multiple recommended information, and return the data corresponding to the target recommended information to the terminal so that the terminal displays the target recommended information;

[0252] Among them, the shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior, and the deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the degree of interaction between the user and the recommended information represented by the shallow conversion behavior is lower than that of the deep conversion behavior;

[0253] The recommended information achievement rate is used to represent the ratio between the number of recommended information meeting the preset requirements and the number of all previously acquired recommended information, or the ratio between the resource occupancy of the recommended information meeting the preset requirements and the resource consumption of all previously acquired recommended information;

[0254] The recommended information meeting the preset requirements satisfies that the difference between the actual cost and the preset cost of the recommended information is within the threshold.

[0255] The device for delivering recommended information provided by the embodiments of the present invention specifically executes the process of the above method embodiments. For details, please refer to the content of the above embodiments of the method for delivering recommended information, which will not be elaborated here. The device for delivering recommended information provided by the embodiments of the present invention, by receiving a request for pulling recommended information from a terminal, obtains at least one recommended information corresponding to the request for pulling recommended information, and determines the shallow conversion behavior parameters and deep conversion behavior parameters of the recommended information. Compared with the prior art that only estimates the amount of resource acquisition based on the shallow conversion behavior parameters, by considering the deep conversion behavior, it can more accurately evaluate the delivery effect of the recommended information, and the estimated amount of resource acquisition of the recommended information obtained thereby has the premise that the achievement rate of the recommended information is within a preset range, which is in line with the long-term interests of the recommended information alliance and the recommender. While the estimated amount of resource acquisition of the recommended information is more accurate, it can also reduce the occurrence probability of frequent price adjustment, save the display resources of the recommended information, improve the delivery efficiency of the recommender, and reduce the cost of delivering the recommended information.

[0256] Based on the above embodiments, as an optional embodiment, the shallow conversion behavior includes at least one of exposure, click, and activation;

[0257] The deep conversion behavior includes at least one of next-day retention, three-day retention, and seven-day retention.

[0258] Based on the above embodiments, as an optional embodiment, the estimated resource acquisition amount module includes:

[0259] The shallow target resource acquisition amount unit is used to calculate the shallow target resource acquisition amount according to the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate of the recommended information;

[0260] The floating ratio calculation unit is used to determine the floating ratio of the shallow target resource acquisition amount when the achievement rate of the recommended information meets the preset conditions according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate;

[0261] The estimated resource acquisition amount calculation unit is used to calculate the estimated resource acquisition amount of the recommended information according to the shallow target resource acquisition amount and the floating ratio.

[0262] Based on the above embodiments, as an optional embodiment, the floating ratio calculation unit includes:

[0263] The first coefficient calculation sub-unit is used to obtain the first quotient value of the first ratio and the first calibration coefficient, and obtain the first coefficient according to the difference between the first quotient value and the preset reference value;

[0264] The floating value calculation sub-unit is used to obtain the second quotient value of the second ratio and the second calibration coefficient, determine the difference between the second quotient value and the preset reference value, and obtain the floating value of the second coefficient in the current cycle compared with the second coefficient in the previous cycle according to the product of the difference and the preset step size;

[0265] The second coefficient calculation sub-unit obtains the second coefficient in the previous cycle and combines the floating value to obtain the second coefficient in the current cycle;

[0266] The coefficient multiplication sub-unit is used to obtain the floating ratio according to the first coefficient and the second coefficient in the current cycle.

[0267] Based on the above embodiments, as an optional embodiment, the floating ratio calculation unit further includes:

[0268] The relationship determination sub-unit is used to determine the size relationship between the recommended preset cost and the initial preset cost of the recommender according to the size relationship between the first coefficient and the preset threshold;

[0269] The cost increase notification unit is used to send a cost increase notification to the recommender corresponding to the recommendation information if the recommended preset cost is higher than the initial preset cost of the recommender;

[0270] The floating value calculation sub-unit is specifically used to: if the cancellation recommendation instruction sent by the recommender according to the cost increase notification is not received, obtain the second quotient value of the second ratio and the second calibration coefficient.

[0271] Based on the above embodiments, as an optional embodiment, the first coefficient calculation sub-unit further includes a calibration coefficient determination sub-unit for obtaining the first calibration coefficient. Specifically, the calibration coefficient determination sub-unit includes:

[0272] The actual cost per action acquisition sub-unit is used to acquire the actual cost per action of the recommendation information in the current cycle;

[0273] The deviation calculation sub-unit is used to obtain the first calibration coefficient according to the deviation between the actual cost per action of the recommendation information in the current cycle and the target cost per action.

[0274] Based on the above embodiments, as an optional embodiment, the request acquisition module includes an actual deep conversion rate acquisition sub-module for acquiring the actual deep conversion rate. The actual deep conversion rate acquisition sub-module includes:

[0275] The first parameter acquisition unit is used to acquire the estimated deep conversion rate of the recommendation information in the preset time window as the first parameter;

[0276] A second parameter acquisition unit, configured to acquire the deep conversion number and the activation return ratio of the recommendation information in a preset time window, and obtain a second parameter according to the conversion number and the activation return ratio of the recommendation information in the preset time window;

[0277] A deep conversion rate calculation unit, configured to calculate an actual deep conversion rate according to the first parameter and the second parameter.

[0278] In an embodiment of the present application, an electronic device is provided. The electronic device includes: a memory and a processor; at least one program, stored in the memory, and when executed by the processor, compared with the prior art, by receiving a recommendation information pull request from a terminal, obtaining at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request, determining the shallow conversion behavior parameter and the deep conversion behavior parameter of the recommendation information. Compared with the prior art that only estimates the resource acquisition amount based on the shallow conversion behavior parameter, by considering the deep conversion behavior, the placement effect of the recommendation information can be evaluated more accurately, and the estimated resource acquisition amount of the recommendation information obtained thereby has the premise that the recommendation information achievement rate is within a preset range, which is in line with the long-term interests of the recommendation information alliance and the recommender. While the estimated resource acquisition amount of the recommendation information is more accurate, the occurrence probability of frequent price adjustment can be reduced, the recommendation information display resources can be saved, the placement efficiency of the recommender can be improved, and the placement cost of the recommendation information can be reduced.

[0279] In an alternative embodiment, an electronic device is provided, as Figure 8 shown Figure 8 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. Among them, the processor 4001 and the memory 4003 are connected, such as connected through a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation to the embodiments of the present application.

[0280] The processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of this application. The processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0281] The bus 4002 may include a path for transmitting information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 8 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0282] The memory 4003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0283] The memory 4003 is used to store the application program code for executing the solution of this application, and is controlled by the processor 4001 to execute. The processor 4001 is used to execute the application program code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0284] An embodiment of this application provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When it runs on a computer, the computer can execute the corresponding content in the foregoing method embodiments. Compared with the prior art, by receiving a recommendation information pulling request from a terminal, obtaining at least one piece of recommendation information corresponding to the recommendation information pulling request according to the recommendation information pulling request, and determining the shallow conversion behavior parameter and the deep conversion behavior parameter of the recommendation information. Compared with the prior art that only calculates the estimated resource acquisition amount based on the shallow conversion behavior parameter, by considering the deep conversion behavior, the placement effect of the recommendation information can be evaluated more accurately, and the estimated resource acquisition amount of the recommendation information obtained thereby has the premise that the recommendation information achievement rate is within a preset range, which is in line with the long-term interests of the recommendation information alliance and the recommender. While the estimated resource acquisition amount of the recommendation information is more accurate, it can also reduce the occurrence probability of frequent price adjustment, save the recommendation information display resources, improve the placement efficiency of the recommender, and reduce the placement cost of the recommendation information.

[0285] An embodiment of this application provides a computer program. The computer program includes computer instructions. The computer instructions are stored in a computer-readable storage medium. When the processor of a computer device reads the computer instructions from the computer-readable storage medium and the processor executes the computer instructions, the computer device executes the content shown in the foregoing method embodiments. Compared with the prior art, by receiving a recommendation information pulling request from a terminal, obtaining at least one piece of recommendation information corresponding to the recommendation information pulling request according to the recommendation information pulling request, and determining the shallow conversion behavior parameter and the deep conversion behavior parameter of the recommendation information. Compared with the prior art that only calculates the estimated resource acquisition amount based on the shallow conversion behavior parameter, by considering the deep conversion behavior, the placement effect of the recommendation information can be evaluated more accurately, and the estimated resource acquisition amount of the recommendation information obtained thereby has the premise that the recommendation information achievement rate is within a preset range, which is in line with the long-term interests of the recommendation information alliance and the recommender. While the estimated resource acquisition amount of the recommendation information is more accurate, it can also reduce the occurrence probability of frequent price adjustment, save the recommendation information display resources, improve the placement efficiency of the recommender, and reduce the placement cost of the recommendation information.

[0286] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially according to the indication of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear indication in this document, there is no strict order restriction for the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments, and their execution order is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0287] The above are only some embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for delivering recommended information, characterized in that, Including: Receiving a recommendation information pulling request from a terminal, and obtaining at least one piece of recommendation information corresponding to the recommendation information pulling request according to the recommendation information pulling request; For each piece of recommendation information in the at least one piece of recommendation information, determining the shallow conversion behavior parameter and the deep conversion behavior parameter of each piece of recommendation information, and obtaining an estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter; the estimated resource acquisition amount is used to evaluate the revenue generated by the recommendation information; Determining a target recommendation information from the at least one piece of recommendation information according to the estimated resource acquisition amount of the at least one piece of recommendation information, and returning the data corresponding to the target recommendation information to the terminal so that the terminal displays the target recommendation information; Wherein, the shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior, and the deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the degree of interaction between the user and the recommendation information represented by the shallow conversion behavior is lower than that of the deep conversion behavior; The recommendation information achievement rate is used to represent the ratio of the number of recommendation information meeting the preset requirements to the number of all pre-obtained recommendation information or the ratio of the resource occupancy of the recommendation information meeting the preset requirements to the resource consumption of all pre-obtained recommendation information; the resource occupancy is the total cost of the recommendation information, and the resource refers to at least one of money, human resources, and time, and the resource consumption is the customer acquisition cost of the recommender; The recommendation information meeting the preset requirements satisfies that the difference between the actual cost and the preset cost of the recommendation information meeting the preset requirements is within the threshold; Wherein, the obtaining the estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter includes: Calculating a shallow target resource acquisition amount according to the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate of the recommendation information; Determining a floating ratio of the shallow target resource acquisition amount when the recommendation information achievement rate is within a preset range according to a first ratio of the estimated deep conversion rate to the target deep conversion rate and a second ratio of the actual deep conversion rate to the target deep conversion rate; Calculating the estimated resource acquisition amount of the recommendation information according to the shallow target resource acquisition amount and the floating ratio.

2. The method for delivering recommended information according to claim 1, characterized in that The shallow conversion behavior parameter includes at least one of the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate; the shallow conversion behavior includes at least one of exposure, click, and activation; The deep conversion behavior parameter includes at least one of the target deep conversion rate, the actual deep conversion rate, and the estimated deep conversion rate; the deep conversion behavior includes at least one of next-day retention, three-day retention, and seven-day retention.

3. The method for delivering recommended information according to claim 1, wherein The determining the floating ratio of the shallow target resource acquisition amount when the recommendation information achievement rate meets the preset conditions according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate includes: Obtain the first quotient value of the first ratio and the first calibration coefficient, and obtain the first coefficient according to the difference between the first quotient value and the preset reference value; Obtain the second quotient value of the second ratio and the second calibration coefficient, determine the difference between the second quotient value and the preset reference value, and obtain the floating value of the second coefficient in the current period compared with the second coefficient in the previous period according to the product of the difference and the preset step size; Obtain the second coefficient in the previous period, and obtain the second coefficient in the current period by combining the floating value; Obtain the floating ratio according to the first coefficient and the second coefficient in the current period.

4. The method for delivering recommended information according to claim 3, wherein Before obtaining the second quotient value of the second ratio and the second calibration coefficient, it further includes: Determine the size relationship between the preset cost of the recommended recommendation information and the initial preset cost of the recommender according to the size relationship between the first coefficient and the preset threshold; If the preset cost of the recommended recommendation information is higher than the initial preset cost of the recommender, send a cost increase notice to the recommender corresponding to the recommendation information; Obtaining the second quotient value of the second ratio and the second calibration coefficient includes: If the cancellation recommendation instruction sent by the recommender according to the cost increase notice is not received, obtain the second quotient value of the second ratio and the second calibration coefficient.

5. The method for delivering recommended information according to claim 3, wherein Obtaining the first calibration coefficient includes: Obtain the actual cost per action of the recommendation information in the current period; Obtain the first calibration coefficient according to the deviation between the actual cost per action of the recommendation information in the current period and the target cost per action.

6. The method for delivering recommended information according to claim 1, wherein, Obtaining the actual deep conversion rate includes: Obtain the estimated deep conversion rate of the recommendation information in the preset time window as the first parameter; Obtain the deep conversion number and the activation return ratio of the recommendation information in the preset time window, and obtain the second parameter according to the deep conversion number and the activation return ratio of the recommendation information in the preset time window; Calculate the actual deep conversion rate according to the first parameter and the second parameter.

7. A device for delivering recommended information, characterized in that, It includes: A request acquisition module, configured to receive a recommendation information pull request from a terminal, and obtain at least one recommendation information corresponding to the recommendation information pull request according to the recommendation information pull request; An estimated resource acquisition amount module, configured to determine, for each of the at least one recommendation information, the shallow conversion behavior parameter and the deep conversion behavior parameter of each recommendation information, and obtain the estimated resource acquisition amount when the recommendation information achievement rate is within a preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter; the estimated resource acquisition amount is used to evaluate the revenue generated by the recommendation information; An information return module, configured to determine the target recommendation information from the at least one recommendation information according to the estimated resource acquisition amount of the at least one recommendation information, and return the data corresponding to the target recommendation information to the terminal, so that the terminal displays the target recommendation information; Among them, the shallow conversion behavior parameter represents a parameter related to the shallow conversion behavior, and the deep conversion behavior parameter represents a parameter related to the deep conversion behavior; the degree of interaction between the user and the recommended information represented by the shallow conversion behavior is lower than that represented by the deep conversion behavior. The recommended information achievement rate is used to characterize the ratio between the number of recommended information that meets the preset requirements and the number of all pre-obtained recommended information, or the ratio between the resource occupancy of the recommended information that meets the preset requirements and the resource consumption of all pre-obtained recommended information; the resource occupancy is the total cost of the recommended information, and the resource refers to at least one of money, human resources, and time, and the resource consumption is the customer acquisition cost of the recommender. The recommended information that meets the preset requirements satisfies that the difference between the actual cost and the preset cost of the recommended information that meets the preset requirements is within the threshold. Among them, the estimated resource acquisition amount module obtains the estimated resource acquisition amount when the recommended information achievement rate is within the preset range according to the shallow conversion behavior parameter and the deep conversion behavior parameter, including: Calculating the shallow target resource acquisition amount according to the target cost per action, the estimated click-through rate, and the estimated shallow conversion rate of the recommended information. Determining the floating ratio of the shallow target resource acquisition amount when the recommended information achievement rate is within the preset range according to the first ratio of the estimated deep conversion rate to the target deep conversion rate and the second ratio of the actual deep conversion rate to the target deep conversion rate. Calculating the estimated resource acquisition amount of the recommended information according to the shallow target resource acquisition amount and the floating ratio.

8. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method for delivering the recommended information according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the method for delivering the recommended information according to any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes computer instructions, which are stored in a computer-readable storage medium. When the processor of the computer device reads the computer instructions from the computer-readable storage medium, the processor executes the computer instructions, causing the computer device to execute the steps of the method for delivering the recommended information according to any one of claims 1-6.

Images