Advertisement resource processing method and related device

Abstract

The embodiment of the invention provides an advertisement resource processing method and a related device. The method comprises the following steps: acquiring a commodity candidate set of a specified field; wherein the commodity candidate set comprises commodity information of a plurality of commodities participating in advertisement putting; obtaining target crowd feature data corresponding to the specified field; wherein the target crowd feature data comprises scene crowd feature information corresponding to a plurality of advertisement scenes in a target advertisement platform; and sending advertisement resources formed by the commodity candidate set and the target crowd feature data to the target advertisement platform. The advertisement putting effect can be improved on the basis that an advertisement platform does not need to be improved on a large scale.

Description

Technical Field

[0001] The implementation methods of the present application relate to the field of Internet advertising technology, and more particularly to an advertising resource processing method and related devices. Background Art

[0002] In the related art, in the field of Internet advertising, merchants can place advertisements through advertising platforms. Specifically, the advertising platform can provide a variety of advertising scenarios for placing advertisements, such as pop-up ads, short video ads, or search recommendation ads.

[0003] Usually, advertisers will select the product information that needs to be advertised and provide it to the advertising platform, and the relevant mechanisms of the advertising platform will complete the specific delivery of the advertisement.

[0004] However, the fields targeted by advertising platforms are diverse, and advertising in each field has its own data processing characteristics, making it difficult for advertising platforms to adapt to the differentiated requirements for advertising in various fields. Summary of the invention

[0005] In view of this, multiple embodiments of the present application are dedicated to providing an advertising resource processing method and related devices, which can improve the advertising delivery effect without requiring large-scale improvements to the advertising platform.

[0006] One embodiment of the present application provides an advertising resource processing method, the method comprising: obtaining a product candidate set in a specified field; wherein the product candidate set includes product information of multiple products participating in advertising; obtaining target population characteristic data corresponding to the specified field; wherein the target population characteristic data includes scene population characteristic information corresponding to multiple advertising scenes in a target advertising platform; and sending the advertising resources consisting of the product candidate set and the target population characteristic data to the target advertising platform.

[0007] One embodiment of the present application provides an advertising resource processing device, including: a candidate set acquisition module, used to acquire a commodity candidate set in a specified field; wherein the commodity candidate set includes commodity information of multiple commodities participating in advertising; a feature data acquisition module, used to acquire target population feature data corresponding to the specified field; wherein the target population feature data includes scene population feature information corresponding to multiple advertising scenes in a target advertising platform; a sending module, used to send the advertising resources consisting of the commodity candidate set and the target population feature data to the target advertising platform.

[0008] One embodiment of the present application provides a computer device, which includes a memory and a processor, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to implement the above-mentioned method.

[0009] One embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one computer program, and when the at least one computer program is executed by a processor, the method as described above can be implemented.

[0010] One embodiment of the present application provides a computer program product, which is used to implement the aforementioned method.

[0011] In multiple implementations provided in the present application, by obtaining a product candidate set in a specified field, wherein the product candidate set includes product information of multiple products participating in advertising, and obtaining target population characteristic data corresponding to the specified field, wherein the target population characteristic data includes scene population characteristic information corresponding to multiple advertising scenes in the target advertising platform, the advertising platform can learn the scene population characteristic information adapted to the product candidate set, and thus, can deliver advertisements in scenes suitable for the product information, thereby improving the advertising delivery effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of the module interaction between an advertisement resource processing device and a target advertisement platform provided in one embodiment of the present application.

[0013] Figure 2 A flowchart of an advertising resource processing method provided for one embodiment of the present application.

[0014] Figure 3 A sample schematic diagram is provided for one embodiment of the present application.

[0015] Figure 4 A schematic diagram of a multi-scenario general model provided for one embodiment of the present application.

[0016] Figure 5 A schematic diagram of training a multi-scenario general model is provided for one embodiment of the present application.

[0017] Figure 6 A schematic diagram of a fitting curve provided for one embodiment of the present application.

[0018] Figure 7 A module diagram of an advertising resource processing device provided for one embodiment of the present application.

[0019] Figure 8A schematic diagram of a computer device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0021] In the description of the embodiments of the present application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0022] In the related art, in the field of Internet advertising resource processing, advertising delivery usually requires screening of a candidate set of products in a specified field, and completing the delivery of advertising resources in combination with the characteristic data of the target population. Specifically, in the traditional advertising delivery method, advertisers usually manually screen products from a product set and determine the target population based on simple rules or historical data.

[0023] First, the generation of product candidate sets often relies solely on the basic attributes of products (such as price, inventory, etc.) for screening, lacking a comprehensive assessment of the competitiveness of products in specific advertising scenarios. This simplistic approach makes it difficult to ensure that the generated product candidate sets can accurately match the actual needs of different advertising scenarios, thus limiting the improvement of advertising effectiveness.

[0024] Secondly, in terms of obtaining characteristic data of the target population, related technologies usually rely on static tags or simple behavioral data, and fail to fully integrate the dynamic needs of multi-scenario advertising. There are significant differences in the target populations suitable for different advertising scenarios (such as pop-up ads, short video ads, or search recommendation ads), and a single population characteristic data is difficult to meet these diverse needs, resulting in inefficient integration of advertising resources. Furthermore, in different fields, there will be differences in the applicability of different advertising scenarios. If improvements are made to the advertising platform to adapt to the characteristics of advertising in different fields and scenarios, it will require a very large amount of manpower investment.

[0025] In multiple implementations provided in the present application, the advertising resource processing method may apply an advertising resource processing device. The advertising resource processing device may be an electronic device with certain computing power and network access capabilities. The electronic device may be a desktop computer, a laptop computer, a tablet computer, a smart phone, or a server. The server may also be a distributed server, including multiple processors, memories, network communication modules, etc., which work together to achieve various functions. Alternatively, the server may also be a server cluster formed by several servers, which has higher computing and data processing capabilities. With the development of science and technology, the server may also be implemented using new forms of technical means, such as a new type of "server" based on quantum computing. Of course, in some implementations, the advertising resource processing device may also be a program module running in an electronic device.

[0026] See also Figure 1 . One embodiment of the present application provides an application scenario example of an advertising resource processing device. The advertising resource processing device can provide advertising resources for a target advertising platform by acquiring a product candidate set, target population characteristic data, and generating conversion indicator estimation data, thereby improving the advertising effect without improving the target advertising platform. In this scenario example, taking advertising in the medical and health field as an example, it is specifically explained how the advertising resource processing device processes advertising resources for medical products and target user groups.

[0027] For example, in an advertising scenario in the medical and health field, the advertising resource processing device first obtains a candidate set of medical products from the product data of the medical e-commerce platform. The data in the medical e-commerce platform includes historical advertising records and product information that has not participated in advertising. The advertising resource processing device identifies products with high click-through rates and conversion rates as the first product information by reading historical advertising records, such as a blood pressure monitor with a broad user base. At the same time, for long-tail products that have not participated in advertising, the advertising resource processing device calls the product competitiveness model, combines the product competition factor and the product sales factor, and selects the second product information suitable for advertising promotion. For example, a newly launched smart blood glucose meter is selected into the product candidate set through the screening of the product competitiveness model because of its rapid recent sales growth and high user evaluation. Finally, the advertising resource processing device integrates the first product information (such as a blood pressure monitor) and the second product information (such as a smart blood glucose meter) to form a product candidate set in the health field.

[0028] After completing the construction of the product candidate set, the advertising resource processing device can further obtain the target population feature data. The advertising resource processing device extracts a target sample data set covering multiple advertising scenarios from the user database of the target advertising platform. For example, the first sample data sampled from the natural traffic includes user behavior data that does not participate in the ad click, such as the user's browsing history on the health content page. The second sample data obtained from the advertising traffic includes the behavior data of the user who clicked the ad, such as a user who completed the purchase after clicking the blood pressure monitor ad. The advertising resource processing device combines the natural traffic and advertising traffic data to construct the target sample data set. Further, the advertising resource processing device constructs a first heterogeneous graph and a second heterogeneous graph according to the user feature data, the product feature data and the user behavior data, and calculates the vector similarity of the user feature data in the two. Through the analysis of vector similarity. Finally, the advertising resource processing device identifies the corresponding scene population feature information from the target sample data set. For example, in the short video advertising scenario, the advertising resource processing device identifies the middle-aged and elderly user group who often watches health science popularization short videos as a potential target population.

[0029] The advertising resource processing device can also further generate conversion index estimation data for advertising resources. The advertising resource processing device generates conversion index estimation data related to the advertising delivery effect based on the short video advertising playback data provided by the target advertising platform and the click data of the goods in the product candidate set. For example, by analyzing the short video playback volume, click-through rate and purchase conversion rate of a blood pressure monitor advertisement, the advertising resource processing device generates conversion index estimation data for the product. Specifically, the advertising resource processing device can construct a fitting association relationship between the conversion index estimation data and the time slice. By fitting the association relationship, the advertising resource processing device can predict the performance trend of an advertisement in different time periods, and send the conversion index estimation data to the target advertising platform according to the time slice. For example, the advertising resource processing device recommends to the target advertising platform during the daily peak time period (such as 8 to 10 pm) that more resources should be allocated to the blood pressure monitor advertisement to improve the overall delivery effect.

[0030] In this scenario example, the advertising resource processing device achieves efficient management and optimized delivery of advertising resources through a multi-step optimization process, including the construction of a product candidate set, the extraction of target population feature data, the generation of conversion indicator estimation data, and fitting analysis. By calling a specific product competitiveness model and a user vector similarity analysis model, the advertising resource processing device can adapt to the complex delivery requirements in health advertising scenarios. In this way, the advertising resource processing device not only improves the overall effect of advertising delivery, but also supports the strategy optimization of the target advertising platform in a variety of advertising scenarios.

[0031] Please also read Figure 1 and Figure 2An embodiment of the present application provides an advertisement resource processing method. The advertisement resource processing method may include the following steps.

[0032] Step S110: Acquire a commodity candidate set in a specified field; wherein the commodity candidate set includes commodity information of multiple commodities participating in advertisement delivery.

[0033] Step S120: Acquire target population characteristic data corresponding to the designated field; wherein the target population characteristic data includes scene population characteristic information corresponding to a variety of advertising scenes in the target advertising platform.

[0034] Step S130: sending the advertisement resources consisting of the commodity candidate set and the target population characteristic data to the target advertisement platform.

[0035] In this embodiment, the advertising resource processing device generates advertising resources adapted to the target advertising platform by executing the advertising resource processing method. The advertising resources include a set of commodity candidates and target population feature data. The advertising resource processing device can be used as an independent device or embedded in other electronic devices or servers in the form of a program module.

[0036] In this embodiment, the advertising resource processing device can obtain a set of commodity candidates in a specified field. The commodity candidate set is composed of information on multiple commodities, and is used to represent a set of commodities that can be selected during the advertising delivery process. The commodity information may include but is not limited to basic attributes such as the name, price, inventory, sales record, and user evaluation of the commodity. In some embodiments, the acquisition of the commodity candidate set can be based on existing historical advertising delivery data. For example, the advertising resource processing device can read the historical advertising delivery records stored in the target advertising platform to filter out commodity information with a high click-through rate or conversion rate in the specified field to form a commodity candidate set. In other embodiments, the acquisition of the commodity candidate set can be performed by calling a commodity competitiveness model and filtering the commodity information in combination with the sales volume factor and competition factor of the commodity. For example, in the health field, the commodity candidate set can include information on commodities such as medicines and medical devices, which can be determined by specific screening rules in the health field.

[0037] In this embodiment, on the basis of obtaining the candidate set of commodities, the advertising resource processing device can further obtain target population characteristic data. The target population characteristic data can be used to represent the characteristic information of the target user group related to the advertising resources, especially the scene population characteristic information corresponding to various advertising scenes in the target advertising platform. The scene population characteristic information may include the user's historical behavior data (such as browsing records, click records, purchase records, etc.), interest tags, demographic information (such as age, gender, occupation, etc.) and platform-specific user portrait information. For example, in the short video advertising scenario, the target population characteristic data can focus on the user's short video viewing habits and preferences, while in the search recommendation advertising scenario, the target population characteristic data can pay more attention to the user's keyword search records. In some embodiments, the advertising resource processing device can extract the target population characteristic data related to the field based on the needs of the specified field through the user database of the target advertising platform.

[0038] In this embodiment, the advertising resource processing device can integrate the product candidate set and the target population characteristic data to form advertising resources, and send them to the target advertising platform. In some embodiments, the integration of advertising resources can be adjusted according to the specific needs of the target advertising platform. For example, the advertising resource processing device can match the product candidate set with the population characteristic data of the corresponding scene according to the specific characteristics of various advertising scenes in the target advertising platform to generate scene-specific advertising resources. For example, for short video advertising scenes, the product information in the product candidate set that has a high correlation with the short video content can be matched with the population characteristic data of users who watch such short videos; and for pop-up advertising scenes, the product information with wide applicability in the product candidate set can be selected to match with a wider range of population characteristic data.

[0039] In this embodiment, the advertising resource processing device can also provide an external call interface so that it can interact with the target advertising platform or other devices in different ways. For example, in some embodiments, the advertising resource processing device can be operated as a service module of the target advertising platform, and the advertising platform directly calls the function of obtaining the product candidate set and the target population feature data; in other cases, the advertising resource processing device can be operated as an independent service device, receiving a call request from an external system (such as a merchant management system or a third-party advertising agency platform), completing the generation of advertising resources and sending them back to the caller.

[0040] In this embodiment, by obtaining the commodity candidate set and target population characteristic data in the specified field, the advertising resource processing device can adapt to the needs of various advertising scenarios in the target advertising platform, efficiently generate advertising resources, and thus improve the overall effect of advertising delivery. In addition, there is no need to make large-scale improvements to the advertising platform itself, which can improve the advertising delivery effect without changing the advertising platform.

[0041] In some embodiments, the advertising resource processing device can read the first product information corresponding to the advertising delivery plan from the designated product set; wherein the designated product set includes multiple product information; the product information in the designated product set that does not correspond to the advertising delivery plan is input into the product competitiveness model to screen out the second product information suitable for promotion through advertising; wherein the product competitiveness model screens the product information based on the product competition factor and the product sales factor; the product competition factor is used to characterize the competitiveness of the product represented by the product information, and the product sales factor is used to characterize the sales volume of the product represented by the product information; the first product information and the second product information are combined into a product candidate set.

[0042] In this embodiment, the advertising resource processing device can read the first product information from the specified product set. The first product information is used to indicate the product information that has participated in the advertising delivery plan. The advertising resource processing device can obtain the first product information by retrieving the historical advertising delivery records of the target advertising platform. For example, the database of the target advertising platform stores historical advertising data related to a specific field. The advertising resource processing device can extract the corresponding information as part of the first product information by screening the products that perform well in these data. Of course, the advertising resource processing device can also record the product information provided to the target advertising platform, or receive external input to obtain the product information corresponding to the advertising delivery plan. In some embodiments, the advertising resource processing device can be applied to an e-commerce platform, and the product information in the specified product set indicates that the product is sold on the e-commerce platform. The corresponding advertising delivery plan is recorded in the e-commerce platform corresponding to the product information.

[0043] In this embodiment, the advertising resource processing device calls the product competitiveness model to filter the product information in the specified product set that does not correspond to the advertising plan to generate the second product information. The second product information is used to represent a set of products that are suitable for promotion through advertising. The product competitiveness model filters the product information through the product competition factor and the product sales factor. The product competition factor is used to characterize the competitiveness of the product represented in the product information, and measures the performance of the product in the advertising scenario by analyzing the characteristics of the product such as the click-through rate, conversion rate and winning rate. The product sales factor characterizes the sales volume of the product represented in the product information, and is quantified by calculating the sales volume ratio of the product (i.e., the sales volume of the product divided by the overall sales volume of the category to which it belongs), and is used to balance the bias of the product competition factor towards products with less performance data.

[0044] In this embodiment, the commodity competitiveness model is constructed based on the multi-dimensional features of the commodity. The multi-dimensional features may include the price features of the commodity (such as the historical average transaction price, activity price, etc.), the historical statistical features of the commodity (such as click UV, purchase UV, transaction UV, CTR, ATC, CVR, etc.), the features of the category to which the commodity belongs (such as click UV and conversion rate, etc.), and the features of the store to which the commodity belongs (such as CTR and logistics service indicators, etc.). The advertising resource processing device calls the commodity competitiveness model and combines the above features to filter out commodity information that meets the advertising delivery requirements from the specified commodity set, and uses this commodity information as part of the second commodity information.

[0045] In this embodiment, the advertising resource processing device integrates the first product information with the second product information to form a product candidate set. The product candidate set includes both product information that has participated in the advertising plan and product information that is suitable for advertising promotion that is screened out by the product competitiveness model. For example, in an advertising scenario in the health field, the product candidate set may include information on products that have been released, such as medical devices and medicines, and information on high-potential products that are screened out by the competitiveness model. By integrating the first product information and the second product information, the advertising resource processing device can generate a comprehensive product candidate set.

[0046] In this embodiment, by reading the first product information from the specified product set and calling the product competitiveness model to screen the second product information, the advertising resource processing device can efficiently generate a product candidate set that is adapted to the needs of the target advertising platform, thereby improving the efficiency of advertising resource generation and the overall effect of advertising delivery.

[0047] In some embodiments, the advertising resource processing device can obtain a target sample data set covering multiple advertising scenarios in the advertising delivery platform; wherein the target sample data set includes multiple target sample data; the target sample data includes user feature data and product feature data, as well as user behavior data that characterizes the relationship between the user feature data and the product feature data; and identify corresponding scene population feature information corresponding to multiple advertising scenarios from the target sample data set.

[0048] In this embodiment, the target population characteristic data is used to characterize the user group characteristics in the advertising resources that are related to various advertising scenarios of the target advertising platform.

[0049] In this embodiment, the advertising resource processing device constructs target population characteristic data by acquiring a target sample data set. The target sample data set covers a variety of advertising scenarios in the advertising delivery platform and is composed of multiple target sample data. The target sample data includes user characteristic data, product characteristic data, and user behavior data. User characteristic data represents attribute information related to the user, such as age, gender, occupation, interest tags, etc. Product characteristic data can be used to represent product information, such as product name, price, sales volume, evaluation, etc. User behavior data can be used to characterize the interaction between users and products, such as click records, browsing records, purchase records, etc.

[0050] In this embodiment, the advertising resource processing device can identify the scene population characteristic information corresponding to various advertising scenes in the advertising platform by analyzing the target sample data in the target sample data set. The scene population characteristic information specifically reflects the behavior patterns and characteristic distribution of the target user group in different advertising scenes. For example, in the short video advertising scene, the scene population characteristic information may include the short video viewing habits, likes and comments, sharing records, etc. of the target user; while in the search recommendation advertising scene, the scene population characteristic information may more reflect the user's search keyword preference and click behavior.

[0051] In this embodiment, in order to obtain the target sample data set, the advertising resource processing device can achieve this by integrating data from different sources. For example, the advertising resource processing device can sample the first sample data that is unrelated to the advertisement from the natural traffic set, and the natural traffic set includes the interactive behavior records of users in the target advertising platform when they are not affected by the advertisement. The advertising resource processing device can also obtain the second sample data formed based on the advertisement delivery from the advertising traffic set, and the advertising traffic set includes the interactive behavior records generated by the user after being exposed to the advertisement. By combining the first sample data and the second sample data, the advertising resource processing device can generate a target sample data set that comprehensively covers a variety of advertising scenarios.

[0052] In this embodiment, the advertising resource processing device can identify scene crowd characteristic information corresponding to a variety of advertising scenes by scene division and feature analysis of the target sample data set. Through the construction of scene crowd characteristic information, the advertising resource processing device can provide more accurate user group description information for advertising resource generation, thereby improving the effect of advertising delivery. For example, in the short video advertising scene, the advertising resource processing device can optimize the target users based on the scene crowd characteristic information to ensure that the advertising content can reach a more relevant user group; in the search recommendation advertising scene, the advertising resource processing device can adjust the advertising strategy through the scene crowd characteristic information to make the target advertising platform's advertising delivery timing more in line with the user's interests.

[0053] In this embodiment, by acquiring a target sample data set covering multiple advertising scenarios and identifying scene population characteristic information, the advertising resource processing device can effectively adapt to the multi-scenario requirements of the target advertising platform. After the target advertising platform receives the advertising resources, it can deliver advertisements more appropriately.

[0054] In some embodiments, the advertising resource processing device can sample the first sample data from the natural traffic set in the specified field; wherein the first sample data in the natural traffic set is irrelevant to multiple advertising scenarios; obtain the second sample data from the advertising traffic set in the specified field; wherein the second sample data in the advertising traffic set is formed based on advertising delivery; and derive the target sample data set by combining the first sample data and the second sample data.

[0055] In this embodiment, the advertising resource processing device can obtain the first sample data by sampling from the natural traffic set in the specified field. The natural traffic set can represent a set of interactive behavior data between users and commodities generated by themselves in the target advertising platform without being affected by advertising. The first sample data is obtained by sampling, is random and has nothing to do with a variety of advertising scenarios, such as the user's browsing history, search history or purchase history, which reflect the user's natural behavior. The advertising resource processing device can extract these interactive behavior data from the natural traffic set to represent the target sample in a natural state.

[0056] In this embodiment, the advertising resource processing device can obtain the second sample data from the advertising traffic set in the specified field. The advertising traffic set can represent a set of user interaction behavior data formed based on advertising delivery. For example, the click record, purchase record or conversion behavior generated by the user after being exposed to the advertisement. The second sample data reflects the impact of advertising delivery on user behavior and has characteristics related to the advertising scenario. The advertising resource processing device extracts the interactive behavior data related to the advertising delivery in the target advertising platform by analyzing the advertising traffic set data, which is used to represent the target sample under the advertising delivery state.

[0057] In this embodiment, the advertising resource processing device generates a target sample data set by combining the first sample data and the second sample data. The target sample data set covers diversified data of natural traffic and advertising traffic, and can fully reflect the behavioral characteristics of users in natural states and advertising delivery states. Through this combination, the advertising resource processing device can ensure that the target sample data set contains user behavior data in a variety of advertising scenarios. For example, in advertising scenarios in the medical and health field, the natural traffic set may include records of users browsing drugs or medical devices on e-commerce platforms, while the advertising traffic set may include records of users clicking or purchasing after watching health product advertisements. The advertising resource processing device samples the natural traffic set and analyzes the advertising traffic set, and combines the two to form a target sample data set to support population characteristic analysis in different advertising scenarios.

[0058] In some embodiments, the advertising resource processing device can construct a first heterogeneous graph of the first sample data in the natural traffic set and a second heterogeneous graph of the second sample data in the advertising traffic set according to user feature data, product feature data and user behavior data; calculate the vector similarity between the vector representation of the user feature data in the first heterogeneous graph and the vector representation of the user feature data in the second heterogeneous graph; select first target sample data from the natural traffic set and select second target sample data from the advertising traffic set; wherein the vector similarity between the vector representation of the user feature data in the first target sample data and the vector representation of the user feature data in the second target sample data is greater than a specified similarity threshold.

[0059] In this embodiment, the advertising resource processing device can characterize the first sample data in the natural traffic set and the second sample data in the advertising traffic set by constructing a heterogeneous graph network. The first heterogeneous graph is constructed by the first sample data in the natural traffic set, and is used to represent the association between the user feature data, product feature data, and user behavior data of the user in the natural interaction state; the second heterogeneous graph is constructed by the second sample data in the advertising traffic set, and is used to represent the association between the user feature data, product feature data, and user behavior data of the user under the influence of the advertisement.

[0060] In this embodiment, the advertising resource processing device calculates the vector similarity between the vector representation of the user feature data in the first heterogeneous graph and the vector representation of the user feature data in the second heterogeneous graph. The vector representation is generated by a feature embedding method and can efficiently represent the multi-dimensional information of the user feature data. The calculation of vector similarity can adopt the cosine similarity formula:

[0061]

[0062] in, and They represent the vector representation of user feature data in the first heterogeneous graph and the second heterogeneous graph respectively, "·" represents the vector dot product, Represents the magnitude of a vector.

[0063] In this embodiment, the advertising resource processing device selects the first target sample data from the natural traffic set and the second target sample data from the advertising traffic set according to the vector similarity. The advertising resource processing device sets a specified similarity threshold and selects only the sample data whose similarity is greater than the threshold. For example, if the specified similarity threshold is 0.8, only the samples whose user feature data vector similarity is higher than 0.8 in the first target sample data and the second target sample data are selected. Of course, the specified similarity threshold is not limited to 0.8, and can be set according to the specific situation, and can also be 0.6, 0.7, 0.9, etc.

[0064] In this embodiment, by combining the first target sample data extracted from the first heterogeneous graph and the second target sample data extracted from the second heterogeneous graph, the advertising resource processing device can generate a more accurate target sample data set. The target sample data set covers the behavioral characteristics of users in the natural state and the advertising delivery state, and effectively reduces data noise through heterogeneous graph modeling and vector similarity calculation, thereby improving the accuracy and quality of the target sample data set.

[0065] In this embodiment, by constructing a heterogeneous graph network according to user feature data, product feature data and user behavior data, and filtering target sample data based on vector similarity, the advertising resource processing device can generate a high-quality target sample data set, providing a reliable basis for the extraction of target population feature data and scene population feature information, thereby improving the generation efficiency of advertising resources and the overall effect of advertising delivery.

[0066] In some embodiments, a heterogeneous graph with user feature data, product feature data, and user behavior data may be provided in advance, and the data scale of the heterogeneous graph may be very large. However, for a specified field, it is not necessary to use such a large heterogeneous graph to build a heterogeneous graph network for the instruction field. For example, the specified field may be the medical and health field. Of course, in some embodiments, even the overall heterogeneous graph in the medical and health field is very large. In order to comply with this application, a random walk algorithm can be used to obtain a heterogeneous subgraph in a large heterogeneous graph, and the heterogeneous subgraph can be used to train the heterogeneous graph network.

[0067] The optimization objective of the heterogeneous graph network can be expressed as follows.

[0068]

[0069] Among them, f srepresents the cosine similarity function; q and i p represents a pair of positive samples, that is, there is an edge between the two; i n k represents the kth negative sample.

[0070] In some implementations, a GNN (Graph Neural Network) may be used to construct a heterogeneous graph network. In order to improve the quality of vector representation, k-hop negative samples and structural negative samples with controllable difficulty may be constructed.

[0071] In this implementation, for k-hop negative samples with controllable difficulty, such as Figure 3 a. Taking the target node user as an example, the advertising resource processing device selects the commodity neighbor node item within its k-hop (k≥2) range as a negative sample. By adjusting the size of the parameter k, the difficulty of the negative sample can be flexibly controlled. Specifically, the item node closer to the user is considered to be a more difficult negative sample, while the item node farther away is relatively simple. In this process, the k-hop negative sample ignores the influence of the graph structure, prompting the GNN to focus more on the learning of node attribute information, thereby optimizing the generation quality of the target sample data.

[0072] In this embodiment, for structural negative samples, such as Figure 3 b. The advertising resource processing device can retain the neighbor structure between the target node user and its positive sample item, but replace the positive sample item with a negative sample obtained by global negative sampling. At this time, the generated negative sample subgraph is consistent with the real subgraph in topological structure, and the randomness of global negative sampling is introduced. In this way, the advertising resource processing device enables GNN to pay more attention to the learning of node attribute information under the condition that the positive and negative samples have the same graph structure. In addition, since GNN is prone to over-reliance on graph structure in modeling (such as over-smoothing problem), this structural negative sample can effectively alleviate this problem and improve the diversity and accuracy of target sample data.

[0073] In this embodiment, by combining the above two types of negative sample designs, the advertising resource processing device can optimize the training effect of the GNN model, increase the degree of attention to node attribute information, and thus generate a high-quality target sample data set. At the same time, the introduction of these negative samples can provide a more reliable basis for the subsequent generation of advertising resources and optimization of advertising delivery effects.

[0074] In some embodiments, see Figure 4 , a multi-scenario general model can be constructed to generate characteristic data of the target population.

[0075] In this embodiment, in order to optimize the efficiency of advertising delivery, the advertising resource processing device establishes a multi-scenario general model through multi-domain sample fusion technology when facing complex and diverse advertising scenarios. Specifically, each scenario is not modeled separately, which will significantly increase the computational complexity, nor are all scenarios simply merged and modeled without distinction, which will lead to the loss of difference information between scenarios and affect the advertising delivery effect. To solve the above problems, the advertising resource processing device constructs a multi-scenario general model based on fused samples, and uniformly models the user click-through conversion behavior in multiple scenarios.

[0076] In this embodiment, the advertising resource processing device generates a target sample data set by fusing samples, and uses a gated network architecture as the basic framework of a multi-scenario general model. The multi-scenario general model is constructed based on an expert network, and the design of the gated network can effectively distinguish the feature weights in different scenario tasks. However, in the actual training process, since the model parameters of the click task subnetwork and the conversion task subnetwork tend to be homogenized, the advertising resource processing device balances the model by introducing sequence seeds corresponding to different tasks. Figure 5 As shown in the figure, click task sequence and conversion task sequence. Specifically, the sequence seeds of different tasks can guide the model to distinguish between click tasks and conversion tasks, thereby reducing the impact of parameter homogeneity on model effects. Furthermore, the click task sub-network and the conversion task sub-network can learn a weight or selection function through a gating mechanism to represent the shared input vector.

[0077] In this embodiment, in order to further improve the effect optimization capability of the advertising resource processing device, the loss function design of the multi-scenario general model adopts a multi-task joint optimization method. Specifically, the advertising resource processing device introduces BCE loss and BPR loss simultaneously in the multi-scenario general model training. Among them, BCE loss is used to generate a healthy conversion factor, which can measure the user's conversion tendency in the advertising scenario; and BPR loss, as a pair-wise learning method, can deeply explore the user's interest in different products, and is used to generate target population characteristic data. By jointly optimizing the two loss functions, the advertising resource processing device can not only generate accurate healthy conversion factors, but also mine more personalized target population characteristic data.

[0078] In this embodiment, through the multi-scenario universal model, the advertising resource processing device can efficiently fit the user click and conversion behavior in different scenarios. Furthermore, the advertising resource processing device ensures the adaptability and accuracy of the model by generating fusion samples and introducing task sequence seeds. Ultimately, the model can output healthy conversion factors and intelligent targeted populations, thereby comprehensively optimizing the overall effect of advertising delivery.

[0079] In some embodiments, the advertising resource processing device can generate conversion indicator prediction data based on the media data provided by the target advertising platform and the user behavior data generated by the corresponding product information in the product candidate set; wherein the conversion indicator prediction data is used to characterize the advertising delivery effect of the advertising resources; and the conversion indicator prediction data is sent to the target advertising platform.

[0080] In this embodiment, the advertising resource processing device further generates conversion index estimation data to characterize the advertising effect of the advertising resources on the target advertising platform. The conversion index estimation data is generated by analyzing the media data provided by the target advertising platform and the user behavior data generated by the corresponding product information in the product candidate set. The media data is used to represent the relevant information of the interaction between the advertising resources and the user when they are displayed on the target advertising platform, such as the exposure, click volume and dwell time of the advertisement. The user behavior data includes the interaction records between the user and the product information in the product candidate set, such as click records, purchase records and evaluation records.

[0081] In some cases, the target advertising platform itself will generate conversion indicator data and adjust the specific advertising delivery strategy based on the conversion indicator data. However, the target advertising platform will generate conversion indicator data relatively late, which will cause the target advertising platform to adjust the advertising delivery strategy relatively late. In this embodiment, the advertising resource processing device can dynamically generate conversion indicator estimation data and provide it to the target advertising platform, so that the target advertising platform can adjust the advertising delivery strategy in a relatively timely manner.

[0082] In this embodiment, the advertising resource processing device can generate conversion index estimation data by establishing a conversion index estimation model. The conversion index estimation model can be modeled based on the following formula:

[0083]

[0084] Among them, CTR (click-through rate) is used to measure the attractiveness of advertisements to users, and CVR (conversion rate) is used to measure the impact of advertisements on users' purchasing behavior. The advertising resource processing device can calculate CTR by analyzing the exposure and click volume in the media data, and calculate CVR by analyzing the click volume and purchase volume in the user behavior data.

[0085] In this embodiment, the conversion index prediction model can also be optimized in combination with the specific features of the advertising scene. For example, for short video advertising scenes, features such as video length and user dwell time can be introduced as model inputs; for search recommendation advertising scenes, features such as user search keyword matching and product ranking can be introduced as model inputs. By combining these specific features, the advertising resource processing device can more accurately predict the conversion effect of advertising resources in different advertising scenes.

[0086] In this embodiment, the conversion index estimation data generated by the advertising resource processing device may include multiple indicators. For example, CTR, CVR, user purchase frequency and user retention rate. These indicators can provide advertisers and advertising platforms with a comprehensive evaluation of the effect of advertising resources. For example, in the advertising scenario in the medical and health field, the conversion index estimation data can reflect the advertising conversion effect of medical devices or drugs, including the purchase rate and repeat purchase rate after users click on the advertisement.

[0087] In this embodiment, the advertising resource processing device sends the generated conversion index prediction data to the target advertising platform. The target advertising platform can adjust the advertising resource delivery strategy based on the conversion index prediction data, such as optimizing the arrangement order of the product candidate set or adjusting the matching strategy of the scene crowd feature data. In addition, the target advertising platform can also use the conversion index prediction data to improve the user recommendation effect, such as increasing the exposure frequency for high-conversion rate products and optimizing the advertising content for low-conversion rate products.

[0088] In some embodiments, the advertising resource processing device can construct a fitting correlation relationship between the conversion indicator prediction data and the time slice; wherein the fitting correlation relationship is used to represent the changing trend of the conversion indicator prediction data with the time slice; based on the fitting correlation relationship, the corresponding conversion indicator prediction data is sent to the target advertising platform in the time slice of each day.

[0089] In this embodiment, the advertising resource processing device further constructs a fitting correlation between the conversion index prediction data and the time slice to characterize the change trend of the conversion index prediction data with the time slice. The conversion index prediction data is used to characterize the advertising effect of the advertising resources on the target advertising platform. Its generation process combines the media data provided by the target advertising platform and the user behavior data generated by the product information in the product candidate set, specifically including the user's click, conversion, purchase and other behavior records on the advertising resources. The time slice represents a time unit divided according to a specific time interval within a day, such as hours or minutes.

[0090] In this implementation, see Figure 6 , the fitting relationship can be constructed by statistical methods or machine learning models. In practical applications, for example, the fitting equation provided in this solution is:

[0091] Y=a+b·X+c·X 2

[0092] Where: Y represents the estimated data of the conversion index in time slice X; a, b, c can be fitting parameters. In a specific example, for example, a = 0.116461038961039, b = 0.00955058099794936, c = 0.00190533151059467; the correlation coefficient R2 = 0.992743519822021, indicating a high fitting accuracy.

[0093] In this embodiment, the advertising resource processing device can capture the changing rules of the conversion indicator estimation data in different time slices by fitting the above formula, and provide an accurate prediction of the advertising delivery effect.

[0094] In this embodiment, the advertising resource processing device can send corresponding conversion index estimation data to the target advertising platform in each time slice of each day according to the fitting association relationship. For example, for the advertising delivery of a certain day, the advertising resource processing device can predict the conversion index estimation data of each time slice of the day based on the time slice data of the previous day by fitting the association relationship. In this way, the advertising resource processing device can send refined conversion index estimation data to the target advertising platform according to the time slice to support real-time optimization of the advertising delivery effect.

[0095] In this embodiment, by constructing a fitting correlation between conversion index estimation data and time slices, the advertising resource processing device can send optimized conversion index estimation data to the target advertising platform in units of time slices, thereby supporting the target advertising platform to dynamically adjust the advertising delivery strategy. For example, the target advertising platform can adjust the exposure of advertising resources in a certain time slice or optimize the order of product arrangement based on the conversion index estimation data of the time slice, thereby improving the advertising delivery effect.

[0096] See also Figure 7 An embodiment of the present application further provides an advertisement resource processing device. The advertisement resource processing device may include: a candidate set acquisition module, a feature data acquisition module, and a sending module.

[0097] The candidate set acquisition module is used to acquire a commodity candidate set in a specified field; wherein the commodity candidate set includes commodity information of multiple commodities participating in advertisement delivery.

[0098] The characteristic data acquisition module is used to acquire the characteristic data of the target population corresponding to the specified field; wherein the characteristic data of the target population includes scene population characteristic information corresponding to a variety of advertising scenes in the target advertising platform.

[0099] The sending module is used to send the advertising resources consisting of the commodity candidate set and the target population characteristic data to the target advertising platform.

[0100] In this embodiment, the specific functions and effects achieved by the advertising resource processing device can be explained with reference to other embodiments of the present application and will not be repeated here.

[0101] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the processor implements the aforementioned method.

[0102] The present application also provides a computer program product comprising instructions, and when the computer program product is executed by a processor, the method as described above is implemented.

[0103] See also Figure 8 The present description embodiment may provide a computer device, the computer device comprising: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, the instructions are executed by the one or more processors, so that the one or more processors implement the aforementioned method.

[0104] In some embodiments, the computer device may include a processor, a non-volatile storage medium, an internal memory, a communication interface, a display device, and an input device connected by a system bus. The non-volatile storage medium may store an operating system and related computer programs.

[0105] The user information or user account information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, etc.) involved in multiple implementation methods of the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0106] It should be understood that the specific examples in this article are only intended to help those skilled in the art better understand the embodiments of the present application, rather than to limit the scope of the present invention.

[0107] It can be understood that in the various implementations of the present application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation methods of the present application.

[0108] It can be understood that the various embodiments described in this application can be implemented individually or in combination, and the embodiments of this application are not limited to this.

[0109] Unless otherwise stated, all technical and scientific terms used in the embodiments of the present application have the same meaning as those generally understood by those skilled in the art of the technical field of the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present application. The term "and / or" used in the present application includes any and all combinations of one or more related listed items. The singular forms of "a kind of", "above" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0110] It can be understood that the processor of the embodiment of the present application can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method implementation can be completed by the hardware integrated logic circuit or software instructions in the processor. The above processor can be a general processor, a digital signal processor (DigitalSignal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiment of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined and executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0111] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (programmable ROM, PROM), an erasable programmable read-only memory (erasable PROM, EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory may be a random access memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0112] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0113] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method implementation methods and will not be repeated here.

[0114] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0115] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0116] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0117] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0118] The above is only a specific implementation of the present application, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for processing advertising resources, characterized in that: The method comprises: Acquire a product candidate set in a specified field; wherein the product candidate set includes product information of multiple products participating in advertisement delivery; Acquire target population characteristic data corresponding to the specified field; wherein the target population characteristic data includes scene population characteristic information corresponding to a plurality of advertising scenes in the target advertising platform; The advertisement resources consisting of the commodity candidate set and the target population characteristic data are sent to the target advertisement platform.

2. The method according to claim 1, characterized in that: The steps for obtaining a candidate set of products in a specified field include: Reading first commodity information corresponding to an advertisement delivery plan from a designated commodity set; wherein the designated commodity set includes a plurality of commodity information; Inputting the commodity information in the designated commodity set that does not correspond to the advertising plan into the commodity competitiveness model to screen out the second commodity information suitable for advertising promotion; wherein the commodity competitiveness model screens the commodity information based on the commodity competition factor and the commodity sales factor; the commodity competition factor is used to characterize the competitiveness of the commodity represented by the commodity information, and the commodity sales factor is used to characterize the sales volume of the commodity represented by the commodity information; The first product information and the second product information are combined into a product candidate set.

3. The method according to claim 1, characterized in that The step of obtaining characteristic data of the target population corresponding to the specified field includes: Acquire a target sample data set covering multiple advertising scenarios in the advertising delivery platform; wherein the target sample data set includes multiple target sample data; the target sample data includes user feature data and product feature data, and user behavior data representing the relationship between the user feature data and the product feature data; The corresponding scene crowd feature information is identified from the target sample data set corresponding to a plurality of advertising scenes.

4. The method according to claim 3, characterized in that The steps of obtaining a target sample data set covering multiple advertising scenarios in the advertising delivery platform include: Sampling the natural traffic set in the designated field to obtain first sample data; wherein the first sample data in the natural traffic set is irrelevant to a variety of advertising scenarios; Acquire second sample data from the advertisement traffic set in the designated field; wherein the second sample data in the advertisement traffic set is formed based on advertisement delivery; The target sample data set is obtained by combining the first sample data and the second sample data.

5. The method according to claim 3, characterized in that: The steps of obtaining a target sample data set covering multiple advertising scenarios in the advertising delivery platform include: According to the user feature data, the product feature data and the user behavior data, a first heterogeneous graph of the first sample data in the natural traffic set is constructed, and a second heterogeneous graph of the second sample data in the advertising traffic set is constructed; Calculating vector similarity between a vector representation of user feature data in the first heterogeneous graph and a vector representation of user feature data in the second heterogeneous graph; Select first target sample data from the natural traffic set, and select second target sample data from the advertising traffic set; wherein the vector similarity between the vector representation of user feature data in the first target sample data and the vector representation of user feature data in the second target sample data is greater than a specified similarity threshold.

6. The method according to claim 1, characterized in that The method further comprises: Generate conversion index estimation data based on the media data provided by the target advertising platform and the user behavior data generated by the corresponding product information in the product candidate set; wherein the conversion index estimation data is used to characterize the advertising delivery effect of the advertising resources; The conversion indicator estimation data is sent to the target advertising platform.

7. The method according to claim 6, characterized in that The method further comprises: Constructing a fitting correlation relationship between the estimated conversion index data and the time slice; wherein the fitting correlation relationship is used to represent the changing trend of the estimated conversion index data with the time slice; The step of sending the conversion indicator estimation data to the target advertising platform includes: Based on the fitted association relationship, corresponding conversion indicator estimation data is sent to the target advertising platform at a time slice every day.

8. An advertising resource processing device, characterized in that: include: A candidate set acquisition module is used to acquire a commodity candidate set in a specified field; wherein the commodity candidate set includes commodity information of multiple commodities participating in advertisement delivery; A feature data acquisition module, used to acquire target population feature data corresponding to the specified field; wherein the target population feature data includes scene population feature information corresponding to a plurality of advertising scenes in the target advertising platform; The sending module is used to send the advertising resources consisting of the commodity candidate set and the target population characteristic data to the target advertising platform.

9. A computer device, characterized in that: The computer device includes a memory and a processor, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer program, and when the at least one computer program is executed by a processor, the method according to any one of claims 1 to 7 can be implemented.

11. A computer program product, characterized in that The computer program product is used to implement the method according to any one of claims 1 to 7.

Images