Advertisement aggregation delivery monitoring management optimization method, system, equipment and medium

Abstract

The invention discloses an advertisement aggregation putting monitoring management optimization method and device, equipment and a medium, and relates to the technical field of advertisement putting management. According to the method, an advertisement putting protocol of an advertisement platform is converted into a standard putting protocol; the method comprises the steps of obtaining advertisement data of an advertisement platform, generating and optimizing a putting strategy, performing effect prediction on the optimized putting strategy, and further adjusting the putting strategy. Therefore, through standardized conversion of delivery protocols of different platforms, it is possible to optimize advertisement delivery budget, frequency, advertisement materials and targeted target groups based on delivery data of different platforms, so that an advertiser can be helped to make an advertisement delivery strategy with a better effect, and a better advertisement delivery effect is achieved.

Description

Technical Field

[0001] The present invention relates to the field of advertising delivery management, and in particular to an advertising aggregation delivery monitoring management optimization method, system, equipment and medium. Background Art

[0002] Existing advertising delivery management systems require integration with multiple advertising platforms for ad delivery. Each platform utilizes a different API protocol, authentication mechanism, and parameter specifications. Before placing advertising resources, advertisers must adapt to the protocols of the advertising platforms. In existing technologies, advertisers must develop independent adaptation modules for different platforms and manually configure delivery parameters. The fragmentation of multi-platform protocols leads to repeated development of adaptation modules, long adaptation cycles, and high maintenance costs, thus failing to effectively address the development cost and efficiency bottlenecks caused by multi-platform protocol differences. Advertisers are forced to invest significant resources in protocol adaptation, resulting in a lengthy delivery process and wasted resources, severely restricting the scalability and flexibility of advertising delivery. Furthermore, due to inconsistent protocols, cross-platform data cannot be coordinated in real time, further creating data silos. This leads to incomplete user profiles, lagging cross-platform analysis, static strategies that cannot respond to market changes, and difficulty in achieving cross-platform deduplication of ad delivery frequencies. This hinders advertisers from implementing precise ad delivery and dynamic optimization strategies.

[0003] In summary, existing ad delivery management technologies suffer from low access efficiency, high costs, and poor policy consistency due to the fragmentation of multi-platform protocols, which in turn impacts the overall effectiveness and resource utilization of ad delivery. Therefore, providing an efficient multi-platform protocol adaptation method to reduce access costs, improve configuration efficiency, and achieve unified cross-platform policy management has become a core technical issue that urgently needs to be addressed in this field. Summary of the Invention

[0004] To address these core issues, the present invention proposes a method, system, device, and medium for optimizing the monitoring and management of aggregated advertising delivery. By building a unified protocol conversion middle layer, it enables automated adaptation of multi-platform APIs, dynamic parameter mapping, and consistent policy configuration, significantly reducing access costs, improving delivery efficiency, and laying the foundation for further data collaboration and dynamic optimization.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, a method for optimizing the monitoring and management of advertising aggregation delivery is provided, comprising the following steps:

[0007] Obtain the advertising platform's advertising delivery agreement and convert it into a standard delivery agreement;

[0008] Obtain advertising data from the advertising platform according to standard delivery agreements, including impressions, click-through rates, conversion rates, and cost indicators;

[0009] Inputting the advertising data into a delivery strategy optimization model to generate advertising delivery parameters, wherein the delivery strategy optimization model is a deep learning model trained based on a first sample data set, and the first sample data set includes historical advertising data of the advertising platform;

[0010] Inputting advertisement delivery parameters into an advertisement delivery effect prediction model to obtain advertisement delivery effect prediction data, wherein the advertisement delivery effect prediction model is a deep learning model trained based on a second sample data set, and the second sample data set includes sample advertisement material feature data and user historical behavior data;

[0011] Generate advertising delivery adjustment parameters based on delivery effect prediction data.

[0012] Based on the above content, a method for optimizing the monitoring and management of advertising aggregation delivery is provided, which is to convert the delivery protocols of different advertising platforms into standard delivery protocols, and then obtain the standardized advertising delivery data of each platform through the standard delivery protocol, and then generate delivery parameters through the strategy optimization model according to the advertising delivery data of each platform, and then use the effect prediction model to predict the effect of the generated delivery parameters and dynamically adjust the delivery parameters. This application realizes efficient access to multiple platforms through protocol conversion, and solves the problems of low efficiency, high cost and delayed optimization of advertising delivery in the existing technology based on real-time data monitoring and deep learning model strategy optimization and effect prediction, combined with dynamic parameter adjustment, and realizes automated, precise and real-time advertising delivery management, significantly improving advertisers' delivery revenue and resource utilization.

[0013] In a possible design, obtaining the ad serving protocol of the ad platform and converting the ad serving protocol into a standard serving protocol includes the following steps:

[0014] Establish a rule adaptation library, which includes the mapping relationship between standardized parameters and advertising platform parameters;

[0015] The advertisement delivery protocol of the advertisement platform is analyzed, and the advertisement delivery protocol is converted into a standard delivery protocol including the standardized parameters according to the mapping relationship between the standardized parameters and the advertisement platform parameters.

[0016] Based on the above content, this application realizes automated protocol conversion by establishing a rule adaptation library and mapping relationship, reduces the workload of manual configuration, and improves the system deployment speed and standardization.

[0017] In a possible design, advertising data is input into a delivery strategy optimization model to generate advertising delivery parameters, including the following steps:

[0018] The advertising data is organized into a state vector. The advertising data includes the delivery period, historical conversion data, average bid of competing platforms, user device type, and ad creative type.

[0019] Input the state vector into the pre-trained delivery strategy optimization model and output action instructions, where the action instructions include budget allocation ratio adjustment instructions, bid coefficient adjustment instructions, and frequency control threshold adjustment instructions;

[0020] Generate advertising delivery parameters based on action instructions.

[0021] Based on the above content, this application realizes the refined dynamic adjustment of delivery parameters (such as budget allocation, bid coefficient, frequency control) through the state vector input and action instruction output mechanism, effectively responds to market competition and changes in user behavior, and improves advertising response rate and return on investment.

[0022] In a possible design, the training method of the delivery strategy optimization model is as follows:

[0023] Deploy a global parameter server and multiple worker threads;

[0024] Copy the initial delivery strategy optimization model in the global parameter server to the worker thread, and input the historical advertising data of the advertising platform into the delivery strategy optimization model;

[0025] Generate training action instructions based on the parameters of the initial launch strategy optimization model and obtain the decision trajectory corresponding to the training action instructions;

[0026] Calculate the policy gradient corresponding to the delivery strategy optimization model based on the decision trajectory;

[0027] Aggregate the policy gradients of all delivery strategy optimization models and adjust the parameters of the initial delivery strategy optimization model in the global parameter server.

[0028] Based on the above content, this application adopts a distributed asynchronous gradient descent algorithm to train the model, improve training efficiency, shorten the model iteration cycle, and ensure the real-time and accuracy of the strategy optimization model.

[0029] In one possible design, inputting advertisement delivery parameters into the delivery effect prediction model and obtaining advertisement delivery effect prediction data includes the following steps:

[0030] Calculate the predicted exposure rate of the ad based on the user ID data of the user's historical behavior data;

[0031] Obtain feature data of the advertising creative corresponding to the delivery data, input the feature data of the advertising creative and the user's historical behavior data into a convolutional neural network model to predict the user click-through rate and / or user conversion rate of the advertising creative, wherein the feature data of the advertising creative includes visual features and text features of the advertising creative.

[0032] Based on the above content, this application uses convolutional neural networks to process visual and text features, accurately predict user clicks and conversion behaviors, and optimize frequency based on predicted exposure rate to improve delivery accuracy and resource utilization.

[0033] In a possible design, generating advertising delivery adjustment parameters based on delivery effect prediction data includes the following steps:

[0034] Adjust the frequency of advertising based on the predicted exposure rate of the advertisement;

[0035] Replace creatives based on their predicted click-through rate and / or conversion rate;

[0036] A user interest graph is constructed based on advertising materials with high user click-through rates and / or high user conversion rates, and an extended audience list corresponding to the user interest graph is generated through a similar population expansion algorithm.

[0037] This application dynamically replaces materials and adjusts frequency based on predicted data, responding to changes in delivery effects in real time; by building interest graphs and expanding audience lists, it expands the coverage of high-value users and improves advertising reach.

[0038] In one possible design, a user interest graph is constructed based on high-click-through-rate advertising creatives, and an extended audience list corresponding to the user interest graph is generated using a similar population expansion algorithm. The following steps are included:

[0039] Constructing an initial seed user set based on user behavior data whose historical click-through rate is greater than a first preset value;

[0040] A graph neural network model is used to process the social relationships and behavior logs of the seed user set, and a user interest graph is generated through a node embedding update algorithm.

[0041] Calculating the vector cosine similarity between the target user and the seed user in the interest graph, and screening the target users whose similarity is greater than or equal to a second preset value to generate an expanded audience list;

[0042] Collect user behavior data of users in the expanded audience list and update the node weights of the graph neural network model.

[0043] This application uses graph neural networks to process user social relationships and behavioral data to achieve accurate seed user expansion, and improves the accuracy and sustainability of audience expansion through cosine similarity screening and dynamic weight updates.

[0044] In a second aspect, an advertising aggregation delivery monitoring management optimization system is provided, including:

[0045] The access module is used to obtain the advertising platform's advertising delivery agreement and convert the advertising delivery agreement into a standard delivery agreement;

[0046] The data processing module is used to obtain advertising data from the advertising platform according to the standard delivery agreement. The advertising data includes impressions, click-through rates, conversion rates, and cost indicators;

[0047] a strategy optimization module, configured to input the advertising data into a delivery strategy optimization model to generate advertising delivery parameters, wherein the delivery strategy optimization model is a deep learning model trained based on a first sample data set, wherein the first sample data set includes historical advertising data of the advertising platform;

[0048] An effect prediction module is used to input advertising delivery parameters into a delivery effect prediction model to obtain advertising delivery effect prediction data, where the delivery effect prediction model is a deep learning model trained based on a second sample data set, and the second sample data set includes sample advertising material feature data and user historical behavior data;

[0049] and a delivery parameter adjustment module, which is used to generate advertising delivery adjustment parameters based on delivery effect prediction data.

[0050] In a third aspect, the present invention provides a computer device comprising a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the advertising aggregation delivery monitoring management optimization method as in the first aspect or any possible design in the first aspect.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium having instructions stored thereon. When the instructions are executed on a computer, the advertising aggregation delivery monitoring management optimization method as in the first aspect or any possible design in the first aspect is executed.

[0052] Beneficial effects of the above scheme:

[0053] The present invention creatively provides an optimization method for monitoring and managing advertising aggregation delivery. Specifically, by converting the delivery protocols of different advertising platforms into a standard delivery protocol, the problem of inconsistent protocol formats across different platforms is resolved, the adaptation cost of multi-platform access is reduced, and system compatibility and integration efficiency are improved. Data such as exposure, click-through rate, conversion rate, and cost indicators of advertising platforms are acquired in real time, providing an accurate and comprehensive data foundation for subsequent strategy optimization and supporting dynamic monitoring and real-time adjustment. Delivery parameters are generated based on a deep learning model trained on historical data, enabling automated learning of optimal strategies for advertising delivery, improving the scientific nature and efficiency of resource allocation and reducing the need for manual intervention. A predictive model constructed by combining advertising material features with user behavior data can assess the effects of different delivery parameters in advance, reducing trial and error costs and improving decision-making accuracy. Delivery parameters are dynamically adjusted based on the predicted data to achieve closed-loop optimization, ensuring that advertising delivery is always in the optimal state and maximizing delivery revenue. Compared to existing advertising delivery solutions that rely on manual experience, static strategies, or a single data source, the present invention significantly improves delivery effectiveness and automation through closed-loop optimization with deep learning models and real-time data, addressing pain points such as complex multi-platform adaptation, delayed optimization response, and inaccurate audience coverage, thereby possessing greater practicality and commercial value. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 A flowchart of an advertising aggregation delivery monitoring and management optimization method provided in an embodiment of the present application.

[0056] Figure 2 This is a flow chart of the steps corresponding to step S1 in the embodiment of the present application.

[0057] Figure 3 This is a flow chart of the steps corresponding to step S5 in the embodiment of the present application.

[0058] Figure 4 This is a flow chart of the steps corresponding to step S7 in the embodiment of the present application.

[0059] Figure 5 This is a flow chart of the steps corresponding to step S9 in the embodiment of the present application.

[0060] Figure 6 This is a structural diagram of the advertising aggregation delivery monitoring management optimization system provided in an embodiment of the present application.

[0061] Figure 7 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention will be briefly introduced below in conjunction with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structures of the drawings is only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these embodiments without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0063] It should be understood that although the terms first, second, etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are merely used to distinguish one object from another. For example, a first object can be referred to as a second object, and similarly, a second object can be referred to as a first object without departing from the scope of the exemplary embodiments of the present invention.

[0064] It should be understood that the term "and / or" that may appear in this document is merely a description of the association relationship between associated objects, indicating that there may be three relationships. For example, A and / or B can indicate three situations: A exists alone, B exists alone, or A and B exist at the same time. For another example, A, B and / or C can indicate the existence of any one of A, B and C or any combination of them. The term " / and" that may appear in this document describes another type of association object relationship, indicating that there may be two relationships. For example, A / and B can indicate two situations: A exists alone or A and B exist at the same time. In addition, the character " / " that may appear in this document generally indicates that the previous and next associated objects are in an "or" relationship.

[0065] Example

[0066] like Figure 1 As shown, the first aspect of this embodiment provides an advertisement aggregation delivery monitoring management optimization method, which may include but is not limited to the following steps S1 to S9.

[0067] S1. Obtain the advertising delivery agreement of the advertising platform and convert the advertising delivery agreement into a standard delivery agreement.

[0068] In step S1, since different advertising platforms (such as Tencent Ads, Baidu Promotion Platform, and Big Data Engine) use private protocols (API interface format, data field, encryption method, etc.), it is impossible to uniformly access them. Therefore, specifically, obtaining the advertising platform's advertising delivery protocol and converting the advertising delivery protocol into a standard delivery protocol includes the following steps:

[0069] S101: Establish a rule adaptation library, wherein the rule adaptation library includes a mapping relationship between the standardized parameters and the advertising platform parameters;

[0070] S103: Analyze the advertisement delivery protocol of the advertisement platform, and convert the advertisement delivery protocol into a standard delivery protocol including the standardized parameters according to the mapping relationship between the standardized parameters and the advertisement platform parameters.

[0071] Establish a protocol parsing module to regularly obtain the API documents and interface call examples of each platform through the marketing interface (Marketing API) of each advertising platform, and combine natural language processing to parse the meaning of protocol fields, extract request parameters (such as ad slot ID, delivery period, bidding strategy) and response data format.

[0072] After that, a protocol conversion rule library is established to predefine the standard protocol format, including mandatory fields (such as ad ID, budget, targeting conditions) and optional extension fields. Protocol conversion rules are established for each advertising platform to map the private fields of each platform to standard protocol fields, and the conversion logic is defined (such as age range conversion, interest tag mapping, etc.). For example:

[0073] Platform A's "user_id" is mapped to a standard protocol "user_unique_id" using an MD5 cryptographic hash.

[0074] Platform B's "bid_strategy" is mapped to the "outbid_coefficient" of the standard protocol, and the conversion formula is defined at the same time: standard bid = platform B bid * coefficient α.

[0075] It should be noted that this application can also monitor the protocol changes of the advertising platform in real time through the API interface, automatically identify new fields, and assist in manual review. For example, when the "audience_engagement_score" field is added to the Ocean Engine, the system automatically interprets its meaning and maps it to the "engagement_level" of the standard protocol.

[0076] S3. According to the standard delivery agreement, obtain the advertising data of the advertising platform, the advertising data including exposure, click-through rate, conversion rate and cost indicators.

[0077] In step S3, since the advertising data of the advertising platform needs to be synchronized in real time, mainly including exposure, click-through rate, conversion rate and cost indicators, the delivery interface of each platform can be called using the standard protocol to support batch creation of advertising plans, setting delivery parameters (such as daily budget, bidding strategy), and obtaining real-time delivery data.

[0078] For example, the Webhook mechanism or real-time API provided by the platform (such as Real-Time Reporting of Baidu Promotion Platform) is used to receive event streams such as exposure, clicks, and conversions; and platform report data (such as cost and conversion rate summary) is pulled periodically (such as every hour).

[0079] It should be noted that in order to ensure data access security for each platform, it is also necessary to maintain the authentication credentials of each platform in the standard delivery protocol. If the platform interface returns an error code, it will automatically retry or feedback will trigger an alarm.

[0080] After obtaining the ad data, metrics such as impressions, click-through rate, conversion cost, and conversion value need to be calculated in real time. Batch statistics also need to be performed according to a preset window, such as using a sliding window to calculate data for the past hour or day.

[0081] In step S3, considering that there are some abnormal data or some duplicate data in the acquired advertising data, which will affect the feasibility and correctness of the subsequent analysis, it is preferred that the step of acquiring the advertising data of the advertising platform also includes: performing data cleaning on the acquired advertising data, including filtering invalid data (such as duplicate events, abnormal values), standardizing data formats (such as unified timestamp format, currency units), and storing the acquired advertising data in a distributed database or real-time data warehouse to support high-concurrency reading and writing.

[0082] Through the above scheme, this application can realize cross-platform protocol parsing of different advertising platforms, thereby generating a standardized advertising delivery protocol, which facilitates the subsequent comparison of the delivery effects of different advertising platforms.

[0083] S5. Input the advertising data into a delivery strategy optimization model to generate advertising delivery parameters. The delivery strategy optimization model is a deep learning model trained based on a first sample data set, and the first sample data set includes historical advertising data of the advertising platform.

[0084] In the step S5, specifically, the advertisement data is input into a delivery strategy optimization model to generate advertisement delivery parameters, including but not limited to the following steps S501 to S505.

[0085] S501. The advertisement data is formed into a state vector, wherein the advertisement data includes a delivery period, historical conversion data, an average bid of a competing platform, a user device type, and an advertisement material type.

[0086] In step S501, since the acquired advertising data has been processed and invalid data has been filtered out, there may be missing data. Therefore, the missing data can be filled using the mean of the front and back windows. After that, feature extraction is performed on the advertising data. For example, the advertising data can be obtained including the delivery period t, the historical k-period conversion rate data C_t-k, the average bid of the competing platform B_c, the user device type D, and the advertising material type M. The above feature values ​​are combined to form a state vector S = (t, C_t-k, B_c, D, M).

[0087] S503: Input the state vector into the pre-trained delivery strategy optimization model and output action instructions, wherein the action instructions include a budget allocation ratio adjustment instruction, a bid coefficient adjustment instruction, a frequency control threshold adjustment instruction, and corresponding adjustment parameters;

[0088] Exemplarily, in step S503, the state vector S obtained in the above step is input into a pre-trained Asynchronous Advantage Actor-Critic (A3C) model, and the output includes four-dimensional action instructions:

[0089] Budget allocation ratio adjustment parameter ΔP∈[-0.1,0.1]

[0090] Bid coefficient adjustment parameter ΔBid∈[-0.05,0.05]

[0091] Frequency control threshold adjustment parameter ΔFreq∈{-1,0,1}

[0092] Audience expansion coefficient adjustment parameter ΔAud∈{0.8,1.0,1.2}.

[0093] S505: Generate advertisement delivery parameters according to the action instruction.

[0094] Specifically, the budget allocation, advertising frequency, etc. for advertising can be adjusted according to the four-dimensional action instructions output in the above steps.

[0095] The specific training steps of the delivery strategy optimization model in step S5 are as follows:

[0096] Deploy a global parameter server and multiple worker threads;

[0097] Copying the initial delivery strategy optimization model in the global parameter server to the working thread, and inputting the historical advertising data of the advertising platform into the delivery strategy optimization model;

[0098] Generate training action instructions based on the parameters of the initial delivery strategy optimization model, and obtain decision trajectories corresponding to the training action instructions;

[0099] Calculating the policy gradient corresponding to the delivery strategy optimization model according to the decision trajectory;

[0100] Aggregate the policy gradients of all the delivery strategy optimization models and adjust the parameters of the initial delivery strategy optimization model in the global parameter server.

[0101] It should be noted that in this embodiment, a global parameter server and N worker threads are deployed. Each thread replicates the initial model θ. Multithreading processes heterogeneous platform data in parallel, and asynchronous gradient updates reduce communication overhead, improving training efficiency. The historical advertising data is split into training, validation, and test sets in a proportional manner (typically a ratio of 7:2:1).

[0102] Each thread performs the following steps independently:

[0103] Execute the ε-greedy strategy for the current state S_t: randomly select an action with probability ε, otherwise select the action with the largest Q value in the current state;

[0104] Record the decision trajectory data corresponding to the current state, which includes the state vector, action vector, and reward vector;

[0105] Based on the decision trajectory data collected by each thread, the actor network and the critic network are updated, policy gradients are calculated, and the temporal difference error is minimized. The actor network (Actor) calculates policy gradients based on the action sequence in the trajectory to improve decisions; the critic network (Critic) uses the state-reward sequence to evaluate the state value and update the value function.

[0106] After completing the update of K trajectory data, the policy gradient of each thread is pushed to the global server to update the parameters of the delivery strategy optimization model.

[0107] Through the above method, advertisers can evaluate the advertising effectiveness of each different advertising platform and make targeted adjustments to their advertising strategies.

[0108] S7. Input the advertisement delivery parameters into a delivery effect prediction model to obtain delivery effect prediction data of the advertisement. The delivery effect prediction model is a deep learning model trained based on a second sample data set, and the second sample data set includes sample advertisement material feature data and user historical behavior data.

[0109] Specifically, inputting the advertisement delivery parameters into the delivery effect prediction model to obtain the advertisement delivery effect prediction data includes the following steps:

[0110] S701. Calculate the predicted exposure rate of the advertisement based on the user ID data of the user's historical behavior data;

[0111] In the above step S701, the user's historical behavior log is extracted to obtain the user ID, the user's historical exposure times, and the user's historical click times; and the basic information of the ad space is also obtained;

[0112] Calculate the historical click weight of each user. That is, the higher the user's click rate, the greater the historical click weight. Combined with the basic information of the ad position, predict the probability of the user seeing the ad and generate the user's exposure probability score for the ad position.

[0113] S703. Obtain feature data of the advertising material corresponding to the delivery data, and input the feature data of the advertising material and the user's historical behavior data into a convolutional neural network model to predict the user click-through rate and / or user conversion rate of the advertising material, wherein the feature data of the advertising material includes visual features and text features of the advertising material.

[0114] In the above step S703, the feature data of the advertising material includes the visual features and text features of the advertising material. These features can be obtained by manual extraction, or by analyzing the advertising image using a convolutional neural network (extracting features such as color and objects), or by parsing the advertising copy (keywords, emotional tendencies) through natural language processing and embedding the user's historical click records to form a feature vector.

[0115] The three types of features are then input into a fully connected neural network, and the output layer of the neural network generates click-through rate / conversion rate probabilities through the Sigmoid function.

[0116] S9. Generate advertisement delivery adjustment parameters based on the delivery effect prediction data.

[0117] Specifically, generating advertisement delivery adjustment parameters based on delivery effect prediction data includes:

[0118] S901. Adjust the frequency of advertisement delivery according to the predicted exposure rate of the advertisement.

[0119] Specifically, ads with a predicted exposure rate below 40% can be incrementally delivered, increasing their daily frequency or changing their placement. For ads with an exposure rate above 80%, precise control can be implemented by setting a maximum daily exposure for a single user or using a frequency decay algorithm to reduce the exposure frequency for that individual user, freeing up the freed-up time slots and placements for other ads. Furthermore, a 24-hour exposure heat map can be created to screen for high-exposure periods, allowing for better allocation of existing budgets.

[0120] S903: Replace the advertising material according to the expected user click-through rate and / or user conversion rate of the advertising material;

[0121] Specifically, you can determine the benchmark user click-through rate and user conversion rate. When the user click-through rate and user conversion rate of the creative are predicted to be lower than the benchmark multiple times, you can use the creative with a higher predicted user click-through rate or user conversion rate to deliver.

[0122] S905: Construct a user interest graph based on the advertising material with high user click-through rate and / or high user conversion rate, and generate an extended audience list corresponding to the user interest graph through a similar population expansion algorithm.

[0123] Constructing an initial seed user set based on user behavior data whose historical click-through rate is greater than a first preset value;

[0124] Specifically, users can be screened to obtain seed users. Users whose historical click-through rate is greater than 2 times the industry average, or who have conversion behaviors in the past 7 days, or who are highly active in social networking, can be screened as seed users.

[0125] Using a graph neural network model to process the social relationships and behavior logs of the seed user set, and generating a user interest graph through a node embedding update algorithm;

[0126] Specifically, the graph neural network model uses a stochastic gradient descent optimizer, and the loss function includes the click behavior similarity and CTR correlation coefficient.

[0127] Calculating the vector cosine similarity between the target user and the seed user in the interest graph, and screening the target users whose similarity is greater than or equal to a second preset value to generate an expanded audience list;

[0128] Specifically, we obtain the user interest tags (one-hot encoding) and behavior sequences (LSTM encoding) of the seed user and the target user, and use the similarity formula Calculate the cosine similarity between the target user and the seed user, where A k is the weight of the target user in the kth behavior, B k is the weight of the seed user in the kth behavior, Wk is the behavioral importance coefficient, and m represents the number of spatial dimensions where vectors A and B are located.

[0129] Collect user behavior data of users in the expanded audience list and update the node weights of the graph neural network model.

[0130] Specifically, the click / conversion data of the expanded audience is recorded in real time, negative feedback is marked for misjudged samples (high similarity but CTR < 0.5%), and the behavior weight is updated using the following formula:

[0131]

[0132] Among them, Wold is the initial influence weight calculated based on historical behavioral data (such as user activity and conversion rate); Wnew is the dynamic adjustment result of the comprehensive historical weight and real-time feedback; α is the attenuation coefficient; CTRactual is the actual click-through rate of the expanded audience for the advertisement; CTRpredict is the expected click-through rate of the expanded audience for the advertisement.

[0133] This allows you to predict the click-through rate and conversion rate of advertising creatives. For creatives with low click-through rates or conversion rates, you can replace them to improve them. At the same time, you can expand the customer base for advertising, making advertising more targeted to the right audience.

[0134] Therefore, based on the advertising aggregation delivery monitoring and management optimization method described in the above steps S1 to S9, by converting the delivery protocols of different advertising platforms into standard delivery protocols, the problem of inconsistent protocol formats of different platforms is solved, the adaptation cost of multi-platform access is reduced, and the system compatibility and integration efficiency are improved; real-time acquisition of data such as exposure, click-through rate, conversion rate and cost indicators of the advertising platform provides an accurate and comprehensive data foundation for subsequent strategy optimization, supports dynamic monitoring and real-time adjustment; based on the deep learning model trained with historical data, delivery parameters are generated, which can automatically learn the best strategy for advertising delivery and improve the scientificity and efficiency of resource allocation. Reduce the need for manual intervention; by combining the characteristics of advertising materials and user behavior data to build a predictive model, the effects of different delivery parameters can be evaluated in advance, reducing trial and error costs and improving decision-making accuracy; dynamically adjust delivery parameters according to predicted data to achieve closed-loop optimization, ensure that advertising delivery is always in the optimal state, and maximize delivery revenue. Compared with advertising delivery solutions that rely on manual experience, static strategies or a single data source, this invention significantly improves the delivery effect and automation level through deep learning models and real-time data closed-loop optimization, solves pain points such as complex multi-platform adaptation, delayed optimization response, and inaccurate audience coverage, and has higher practicality and commercial value.

[0135] like Figure 2 As shown, the second aspect of this embodiment provides a virtual device for implementing the advertising aggregation delivery monitoring management optimization method described in the first aspect, including an access module, a data processing module, a strategy optimization module, an effect prediction module and a delivery parameter adjustment module that are sequentially connected in communication;

[0136] The access module is used to obtain the advertising delivery agreement of the advertising platform and convert the advertising delivery agreement into a standard delivery agreement;

[0137] A data processing module, configured to obtain advertising data from the advertising platform according to the standard delivery protocol, wherein the advertising data includes exposure, click-through rate, conversion rate, and cost indicators;

[0138] a strategy optimization module, configured to input the advertisement data into a delivery strategy optimization model to generate advertisement delivery parameters, wherein the delivery strategy optimization model is a deep learning model trained based on a first sample data set, wherein the first sample data set includes historical advertisement data of the advertisement platform;

[0139] an effect prediction module, configured to input the advertisement delivery parameters into a delivery effect prediction model to obtain delivery effect prediction data for the advertisement, wherein the delivery effect prediction model is a deep learning model trained based on a second sample data set, wherein the second sample data set includes sample advertisement material feature data and user historical behavior data;

[0140] The delivery parameter adjustment module is used to generate advertisement delivery adjustment parameters according to the delivery effect prediction data.

[0141] The working process, working details and technical effects of the aforementioned device provided in the second aspect of this embodiment can be found in the advertising aggregation delivery monitoring management optimization method described in the first aspect, and will not be repeated here.

[0142] like Figure 3As shown, the third aspect of this embodiment provides a computer device for executing the advertising aggregation delivery monitoring management optimization method as described in the first aspect, including a memory, a processor and a transceiver connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the inflatable cushion sample simulation modeling method as described in the first aspect. For example, the memory may include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a flash memory (Flash Memory), a first-in-first-out memory (FIFO) and / or a first-in-last-out memory (FILO), etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. In addition, the computer device may also include, but is not limited to, a power module, a display screen and other necessary components.

[0143] The working process, working details and technical effects of the aforementioned computer equipment provided in the third aspect of this embodiment can be found in the inflatable cushion sample simulation modeling method described in the first aspect, and will not be repeated here.

[0144] A fourth aspect of this embodiment provides a computer-readable storage medium storing instructions including the method for simulating and modeling an inflatable cushion sample as described in the first aspect. Specifically, the computer-readable storage medium stores instructions that, when executed on a computer, execute the method for simulating and modeling an inflatable cushion sample as described in the first aspect. The computer-readable storage medium refers to a data storage medium and may include, but is not limited to, a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick. The computer may be a general-purpose computer, a dedicated computer, a computer network, or other programmable device.

[0145] The working process, working details and technical effects of the aforementioned computer-readable storage medium provided in the fourth aspect of this embodiment can be referred to the inflatable cushion sample simulation modeling method described in the first aspect, and will not be repeated here.

[0146] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A method for optimizing monitoring and management of advertising aggregation delivery, characterized in that: The following steps are involved: Obtaining the advertising platform's advertising delivery agreement and converting the advertising delivery agreement into a standard delivery agreement; Obtaining advertising data from the advertising platform according to the standard delivery agreement, the advertising data including exposure, click-through rate, conversion rate, and cost indicators; Inputting the advertising data into a delivery strategy optimization model to generate advertising delivery parameters, wherein the delivery strategy optimization model is a deep learning model trained based on a first sample data set, wherein the first sample data set includes historical advertising data of the advertising platform; Inputting the advertisement delivery parameters into a delivery effect prediction model to obtain delivery effect prediction data of the advertisement, wherein the delivery effect prediction model is a deep learning model trained based on a second sample data set, and the second sample data set includes sample advertisement material feature data and user historical behavior data; Generate advertisement delivery adjustment parameters based on the delivery effect prediction data.

2. The method according to claim 1, characterized in that The step of obtaining the advertisement delivery protocol of the advertisement platform and converting the advertisement delivery protocol into a standard delivery protocol includes the following steps: Establishing a rule adaptation library, wherein the rule adaptation library includes a mapping relationship between standardized parameters and advertising platform parameters; The advertisement delivery protocol of the advertisement platform is analyzed, and the advertisement delivery protocol is converted into a standard delivery protocol including the standardized parameters according to a mapping relationship between the standardized parameters and the advertisement platform parameters.

3. The method according to claim 1, characterized in that The step of inputting the advertisement data into a delivery strategy optimization model to generate advertisement delivery parameters comprises the following steps: Combining the advertising data into a state vector, the advertising data including delivery period, historical conversion data, average bid of competing platforms, user device type, and advertising material type; Inputting the state vector into the pre-trained delivery strategy optimization model and outputting action instructions, wherein the action instructions include budget allocation ratio adjustment instructions, bid coefficient adjustment instructions, and frequency control threshold adjustment instructions; Generate advertisement delivery parameters according to the action instruction.

4. The method according to claim 3, characterized in that The training method of the delivery strategy optimization model is as follows: Deploy a global parameter server and multiple worker threads; Copying the initial delivery strategy optimization model in the global parameter server to the working thread, and inputting the historical advertising data of the advertising platform into the delivery strategy optimization model; Generate training action instructions based on the parameters of the initial delivery strategy optimization model, and obtain decision trajectories corresponding to the training action instructions; Calculating the policy gradient corresponding to the delivery strategy optimization model according to the decision trajectory; Aggregate the policy gradients of all the delivery strategy optimization models and adjust the parameters of the initial delivery strategy optimization model in the global parameter server.

5. The method according to claim 1, characterized in that The step of inputting the advertisement delivery parameters into the delivery effect prediction model to obtain the advertisement delivery effect prediction data comprises the following steps: Calculating the predicted exposure rate of the advertisement based on the user ID data of the user's historical behavior data; Obtain characteristic data of the advertising material corresponding to the delivery data, input the characteristic data of the advertising material and user historical behavior data into a convolutional neural network model to predict the user click-through rate and / or user conversion rate of the advertising material, wherein the characteristic data of the advertising material includes visual features and text features of the advertising material.

6. The method according to claim 5, characterized in that Generating the advertisement delivery adjustment parameters according to the delivery effect prediction data includes the following steps: Adjusting the frequency of advertisement delivery according to the predicted exposure rate of the advertisement; replacing the advertising creative according to the expected user click-through rate and / or user conversion rate of the advertising creative; A user interest graph is constructed based on the advertising material with high user click-through rate and / or high user conversion rate, and an extended audience list corresponding to the user interest graph is generated through a similar population expansion algorithm.

7. The method according to claim 5, characterized in that The step of constructing a user interest graph based on the advertisement material with a high click-through rate and generating an extended audience list corresponding to the user interest graph by using a similar group expansion algorithm includes the following steps: Constructing an initial seed user set based on user behavior data whose historical click-through rate is greater than a first preset value; Using a graph neural network model to process the social relationships and behavior logs of the seed user set, and generating a user interest graph through a node embedding update algorithm; Calculating the vector cosine similarity between the target user and the seed user in the interest graph, and screening the target users whose similarity is greater than or equal to a second preset value to generate an expanded audience list; Collect user behavior data of users in the expanded audience list and update the node weights of the graph neural network model.

8. An advertising aggregation delivery monitoring management optimization system, characterized by: include: The access module is used to obtain the advertising delivery agreement of the advertising platform and convert the advertising delivery agreement into a standard delivery agreement; A data processing module, configured to obtain advertising data from the advertising platform according to the standard delivery protocol, wherein the advertising data includes exposure, click-through rate, conversion rate, and cost indicators; a strategy optimization module, configured to input the advertisement data into a delivery strategy optimization model to generate advertisement delivery parameters, wherein the delivery strategy optimization model is a deep learning model trained based on a first sample data set, wherein the first sample data set includes historical advertisement data of the advertisement platform; an effect prediction module, configured to input the advertisement delivery parameters into a delivery effect prediction model to obtain delivery effect prediction data for the advertisement, wherein the delivery effect prediction model is a deep learning model trained based on a second sample data set, wherein the second sample data set includes sample advertisement material feature data and user historical behavior data; and a delivery parameter adjustment module, configured to generate advertisement delivery adjustment parameters based on the delivery effect prediction data.

9. A computer device, characterized in that: It includes a memory, a processor and a transceiver that are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the advertising aggregation delivery monitoring management optimization method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on the computer, the advertising aggregation delivery monitoring management optimization method as described in any one of claims 1 to 7 is executed.

Images