AI marketing advertisement putting optimization system and method fusing multi-dimensional data

Abstract

The invention discloses an AI marketing advertisement putting optimization system and method fusing multi-dimensional data, and relates to the technical field of data processing, and the system comprises a historical record obtaining module which extracts the historical putting record of a target optimization block; the terminal data interaction module interacts with the advertisement putting management platform based on the historical putting record to obtain multi-dimensional terminal data; a grid division module performs grid division on the target optimization block to generate a putting grid and time sequence label data; the mapping model construction module constructs a label mapping model based on the grid time sequence label data, the historical delivery record and the multi-dimensional terminal data; the advertisement task processing module generates an advertisement label cluster by using a label mapping model according to the advertisement task information; and the putting scheme optimization module realizes the optimal matching of the advertisement task and the label cluster through a multi-objective optimization algorithm, and outputs an advertisement putting scheme, thereby achieving the technical effects of improving the data utilization rate, providing predictive putting optimization and improving the putting income.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an AI marketing advertising delivery optimization system and method integrating multi-dimensional data. Background Art

[0002] Advertisement delivery optimization through data-driven means can deliver accurate advertisements at the right time, in the right place, and to the right audience, thereby improving advertising effectiveness and conversion rates. Existing advertising delivery optimization technologies have low data utilization rates for multi-dimensional user data, and are usually single-target based, failing to consider multi-target needs. For example, CN117408749B (A method and system for generating advertising delivery strategies) "establishes a historical advertising revenue data-time mapping relationship and a historical advertising delivery strategy-time mapping relationship" to make two single-target mapping decisions, which has technical problems such as insufficient data analysis depth, lack of delivery predictability, and low delivery revenue. Summary of the invention

[0003] The present invention provides an AI marketing advertising delivery optimization system and method that integrates multi-dimensional data to solve the technical problems of insufficient data analysis depth, lack of delivery predictability, and low delivery returns in the prior art, and to achieve the technical effects of improving data utilization, providing predictive delivery optimization, and increasing delivery returns.

[0004] On the one hand, the present invention provides an AI marketing advertising delivery optimization system integrating multi-dimensional data.

[0005] On the other hand, the present invention also provides an AI marketing advertising optimization method that integrates multi-dimensional data.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] An AI marketing advertising optimization system integrating multi-dimensional data, wherein the system comprises:

[0008] A historical record acquisition module is used to determine the target optimization block and obtain the historical delivery record of the target optimization block.

[0009] The terminal data interaction module is used to interact with the advertising delivery management platform based on the historical delivery records to obtain associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data.

[0010] A grid division module is used to divide the target optimization block into grids, obtain multiple delivery grids, and construct multiple grid time series label clusters of the multiple delivery grids to obtain grid time series label data.

[0011] A mapping model construction module, wherein the mapping model construction module is used to construct a label mapping model based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, wherein the label mapping model includes multiple parallel audience label prediction channels and label mapping decision networks.

[0012] The advertising task processing module is used to obtain the advertising task information of the target optimization block, extract the advertising task form and the task timestamp, and input the task timestamp into the label mapping model to obtain multiple advertising label clusters of multiple delivery grids.

[0013] The delivery plan optimization module is used to perform optimal matching between the advertisement task form and the multiple advertisement tag clusters based on a multi-objective optimization algorithm to obtain an advertisement delivery plan.

[0014] An AI marketing advertising optimization method integrating multi-dimensional data, wherein the method comprises:

[0015] Determine the target optimization block and obtain the historical delivery records of the target optimization block.

[0016] Based on the historical delivery records, the interactive advertisement delivery management platform obtains associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data.

[0017] The target optimization block is grid-divided to obtain a plurality of delivery grids, and a plurality of grid time series label clusters of the plurality of delivery grids are constructed to obtain grid time series label data.

[0018] A label mapping model is constructed based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, wherein the label mapping model includes a plurality of parallel audience label prediction channels and a label mapping decision network.

[0019] The advertisement task information of the target optimization block is obtained, the advertisement task form and the task timestamp are extracted, and the task timestamp is input into the label mapping model to obtain a plurality of advertisement label clusters of the plurality of delivery grids.

[0020] Based on a multi-objective optimization algorithm, the advertising task form is optimally matched with the multiple advertising tag clusters to obtain an advertising delivery plan.

[0021] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0022] The AI ​​marketing advertising delivery optimization system and method for integrating multi-dimensional data of the present invention include: a historical record acquisition module for determining a target optimization block and obtaining its historical delivery record; a terminal data interaction module for interacting with an advertising delivery management platform and obtaining relevant multi-dimensional terminal data according to historical delivery records, including multi-dimensional personal terminal data and multi-dimensional public terminal data; a grid division module for gridding the target optimization block and obtaining grid time series label data; a mapping model construction module for constructing a label mapping model based on grid time series label data, historical delivery records and multi-dimensional terminal data, including multiple parallel audience label prediction channels and label mapping decision networks; an advertising task processing module for extracting advertising task information, task forms and timestamps, and inputting the timestamps into the label mapping model to obtain advertising label groups of multiple delivery grids; and a delivery plan optimization module for matching advertising task forms and multiple advertising label groups through a multi-objective optimization algorithm to obtain the best advertising delivery plan.

[0023] The AI ​​marketing advertising delivery optimization system and method integrating multi-dimensional data of the present invention solves the technical problems of insufficient data analysis depth, lack of delivery predictability, and low delivery returns, and achieves the technical effects of improving data utilization, providing predictive delivery optimization, and increasing delivery returns. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a schematic diagram of the structure of the AI ​​marketing advertising optimization system that integrates multi-dimensional data;

[0025] Figure 2 A flowchart of an AI marketing advertising optimization method that integrates multi-dimensional data.

[0026] Explanation of reference numerals: history record acquisition module 11, terminal data interaction module 12, grid division module 13, mapping model construction module 14, advertising task processing module 15, delivery plan optimization module 16. DETAILED DESCRIPTION

[0027] The technical solution provided in the embodiments of the present invention is to solve the technical problems of insufficient data analysis depth, lack of delivery predictability, and low delivery benefits in the prior art. The overall idea adopted is as follows:

[0028] First, determine the target optimization block and obtain the historical delivery record of the target optimization block; then, based on the historical delivery record, interactive advertising delivery management platform, obtain the associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data; then, grid the target optimization block, obtain multiple delivery grids, and construct multiple grid time series label clusters of multiple delivery grids to obtain grid time series label data; then, construct a label mapping model based on the grid time series label data, the historical delivery record and the multi-dimensional terminal data, wherein the label mapping model includes multiple parallel audience label prediction channels and label mapping decision networks; then, obtain the advertising task information of the target optimization block, extract the advertising task form and the task timestamp, and input the task timestamp into the label mapping model to obtain multiple advertising label clusters of multiple delivery grids; finally, based on the multi-objective optimization algorithm, perform optimal matching between the advertising task form and the multiple advertising label clusters to obtain an advertising delivery plan.

[0029] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0030] Example 1

[0031] Figure 1 The present invention is a schematic diagram of the structure of an AI marketing advertising optimization system integrating multi-dimensional data, wherein the system includes:

[0032] The historical record acquisition module 11 is used to determine the target optimization block and acquire the historical delivery record of the target optimization block.

[0033] Specifically, the target optimization block refers to a specific area where advertising delivery optimization is required. The target optimization block is divided based on the Internet platform or geographic location information. For example, it can be a specific city, business district, administrative division, community, park, CBD, etc. corresponding to a single Internet platform or an area in the Internet platform.

[0034] Specifically, the historical delivery record is the advertising delivery data within the target optimization block, which provides a comprehensive description of the advertising delivery behavior. The historical delivery record contains multiple related advertising campaign contents, specific delivery devices, delivery time and other information. In particular, the above-mentioned specific delivery devices are Internet display terminals or offline devices within the target optimization block, such as advertising spaces on portal websites, video websites, etc., digital advertising screens connected to the Internet, etc.; by obtaining the historical delivery records of the target block, strong data support is provided for precise advertising delivery optimization.

[0035] The terminal data interaction module 12 is used to interact with the advertising delivery management platform based on the historical delivery records to obtain associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data.

[0036] Specifically, through the interactive advertising delivery management platform, multi-dimensional terminal data related to historical delivery records are obtained, including multi-dimensional personal terminal data and multi-dimensional public terminal data, which are used to analyze delivery effects and optimize strategies, thereby improving the accuracy and efficiency of advertising. Among them, multi-dimensional personal terminal data is used to describe the device and behavioral characteristics directly related to the target audience, that is, to reflect the individualized performance of users in advertising interaction; multi-dimensional public terminal data describes the operating status and effects of public terminals involved in advertising delivery, reflecting the breadth of delivery coverage and environmental impact.

[0037] Specifically, multi-dimensional terminal data is obtained through the API or management background of the advertising delivery management platform, where the advertising delivery management platform is a platform used to execute advertising delivery, track advertising effects in real time, and record advertising delivery logs. In other words, the advertising delivery management platform can cover multiple online and offline delivery channels and integrate different terminal and user data.

[0038] In some embodiments, based on the historical delivery records, the interactive advertising delivery management platform obtains the associated multi-dimensional terminal data, and the execution steps include:

[0039] Based on the target optimized block, the corresponding block position information and block window information are obtained; with the block position information and the block window information as call constraints, the advertising delivery management platform is interacted with to extract the audience effectiveness data and audience tag data of multiple personal terminals, and output as the multi-dimensional personal terminal data; based on the historical delivery records, the corresponding multiple public terminals are matched on the advertising delivery management platform, and the public effectiveness data of the multiple public terminals are extracted, and output as the multi-dimensional public terminal data.

[0040] Specifically, first obtain the block position information and block window information of the target optimization block, wherein the block position information is used to represent the logical boundaries and constraints of the target block in the marketing goal, exemplarily including the target audience, delivery channel, device type, geographical distribution, etc. of the advertising, in order to clarify the optimization scope and restriction conditions of the advertising; the block window information defines the sampling time range for obtaining the multi-dimensional terminal data of the target block.

[0041] Specifically, the block position information and block window information of the target optimization block are used as call constraints to filter personal terminal data and obtain multi-dimensional personal terminal data, wherein the personal terminal data includes audience effectiveness data and audience label data. The audience effectiveness data is used to measure the performance of personal terminal users in advertising campaigns, including multi-dimensional performance indicators such as exposure times, click times, conversion times, interaction rates, and conversion rates; the audience label data is used to describe the attribute characteristics of the audience users corresponding to the personal terminal, including multi-dimensional characteristic indicators such as demographic characteristics, interest preferences, geographic location, and device type.

[0042] Exemplarily, the interactive advertising delivery management platform first extracts the corresponding target audience, delivery channel, device type, geographical distribution and other information including the user behavior records of the personal terminal in the block location information, and then filters the extracted audience effectiveness data and label data according to the block window information, and organizes and outputs them into multi-dimensional personal terminal data.

[0043] Specifically, a public terminal refers to an advertising terminal in which the corresponding identification data (such as personal information, device type, geographic coordinates, etc.) in the target block is missing, such as a user terminal that cannot obtain key user data due to the user turning on privacy protection functions (such as private browsing mode, sandbox mode, ad blocking tools, etc.), the user's IP address is hidden or displayed as an anonymous proxy, the cookie information cannot be obtained or the user refuses cookie tracking, data collection is limited, etc.; in other words, a public terminal can be considered as a user terminal with unclear user data; public performance data is used to describe the performance of public terminals in advertising delivery, including, for example, display frequency, display duration, population coverage (the estimated number of viewers who have seen the advertisement), interaction data (user scanning code, clicking on interactive screens, and other behavioral data), and exposure performance (such as the ratio of population coverage to display frequency)

[0044] Exemplarily, the corresponding public terminal is matched in the advertising delivery management platform according to the advertising campaign ID or device ID, and then, based on the same principle of obtaining multi-dimensional personal terminal data as mentioned above, the public terminal data related to the advertising campaign is filtered by region and time and organized into multi-dimensional public terminal data.

[0045] The above process, by obtaining detailed block location information and window information, obtains high-precision multi-dimensional personal terminal data and multi-dimensional public terminal data, which helps to achieve more accurate audience matching and thus improve the relevance of advertising.

[0046] The grid division module 13 is used to divide the target optimization block into grids, obtain multiple delivery grids, and construct multiple grid time series label clusters of the multiple delivery grids to obtain grid time series label data.

[0047] Specifically, the size of the divided grid is determined according to the block characteristics of the target optimization block, and the target block is divided into multiple small grids (i.e., delivery grids) to facilitate more fine-grained analysis and delivery; wherein, multiple delivery grids have corresponding grid timing label clusters, and the grid timing labels are used to reflect the audience characteristics of the delivery grids in different time windows, thereby facilitating more accurate advertising delivery analysis.

[0048] In other words, through grid division, refined management and analysis of target blocks can be achieved. By constructing multiple grid time-series label clusters of multiple delivery grids, audience behaviors and preferences that change over time can be identified, which helps to improve the effectiveness and efficiency of advertising delivery.

[0049] In some embodiments, the target optimization block is gridded, a plurality of delivery grids are obtained, and a plurality of grid time series label clusters of the plurality of delivery grids are constructed to obtain grid time series label data. The execution steps include:

[0050] The grid granularity of the target optimization block is defined, and based on the grid granularity, the target optimization block is divided into multiple delivery grids; multiple groups of multidimensional terminal grid data corresponding to the multiple delivery grids are obtained by matching the multidimensional terminal data; based on the sliding window method, multiple groups of the multidimensional terminal grid data are statistically analyzed, and multiple grid time series label clusters of the multiple delivery grids are extracted, and output as the grid time series label data.

[0051] Specifically, first, according to the analysis requirements and block size, define the appropriate grid granularity to ensure fine-grained analysis of the data, and based on the determined grid granularity, divide the target optimization block into multiple delivery grids, where the grid granularity can be considered as the smallest delivery unit for advertising delivery, such as different delivery platforms (websites, APPs, APIs, etc.), different IP addresses, different language types, different device types (PC, Android, IOS, Linux), etc.; then, according to the grid identity information of multiple delivery grids, match the multi-dimensional terminal data with each delivery grid to generate multiple groups of multi-dimensional terminal grid data corresponding to each grid.

[0052] Specifically, the size and step size of the sliding window are determined to enable continuous analysis in the time dimension, wherein the size of the sliding window determines the representativeness of the analysis at a single moment, and the step size determines the time series sampling rate of the acquired grid time series label cluster. For example, if you need to analyze diurnal changes, you can select an hourly window; if you are concerned about the difference between weekends and weekdays, select a day-level window.

[0053] Specifically, the sliding window method is used to perform statistical analysis on the multi-dimensional terminal data of each delivery grid with a set sliding window size and step size to identify patterns and trends in the data, and then generate a corresponding grid time series label cluster for each delivery grid. The grid time series label cluster can reflect the audience characteristics, behavior patterns, active time and other information of each delivery grid at different times.

[0054] The above steps, through grid division and sliding window analysis, realize fine-grained data analysis of the target block, and then identify and capture the audience behavior and characteristics that change over time through time series label clusters, providing accurate data support for subsequent optimization decisions, helping to optimize advertising delivery decisions and improve delivery effects.

[0055] The mapping model building module 14 is used to build a label mapping model based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, wherein the label mapping model includes multiple parallel audience label prediction channels and label mapping decision networks.

[0056] Specifically, the audience label prediction channel is used to predict the audience characteristics of the corresponding delivery grid according to the timestamp of the target delivery time. The audience characteristics are represented as predicted label clusters, and multiple labels in the predicted label clusters are marked with corresponding confidence levels; the label mapping decision network is used to convert the predicted label clusters into corresponding advertising label clusters, thereby guiding the delivery of advertisements.

[0057] In some embodiments, based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, a label mapping model is constructed, and the execution steps include:

[0058] Statistically analyzing the periodic characteristics of multiple grid time series label clusters in the grid time series label data, and constructing multiple audience label prediction channels; defining a delivery efficiency constraint based on the expected delivery effect of the target optimization block, and performing data cleaning of the historical delivery records and the multidimensional terminal data according to the delivery efficiency constraint to obtain standard delivery records and standard multidimensional terminal data; using the standard multidimensional terminal data as training input data and the standard delivery records as training target output, constructing and training the label mapping decision network; connecting multiple audience label prediction channels in parallel, and generating the label mapping model by connecting the output ends of multiple audience label prediction channels with the input end of the label mapping decision network, wherein multiple parallel audience label prediction channels are used to respectively predict the real-time label clusters of multiple delivery grids.

[0059] Specifically, first, multiple grid time series label clusters in the grid time series label data are respectively input into multiple parallel audience label prediction channels, wherein each audience label prediction channel uses a convolutional neural network (CNN) or a recurrent neural network (RNN) to extract periodic time series features in the grid time series label cluster; then, the expected delivery effect (such as click-through rate, conversion rate) is defined according to the business objectives, and delivery efficiency constraints are defined to ensure that the delivery strategy meets business needs, and according to the delivery efficiency constraints, the historical delivery records and multi-dimensional terminal data are cleaned, and the data that does not meet the delivery efficiency constraints is improved to improve the data quality, and standard delivery records and standard multi-dimensional terminal data are generated as the basis for model training; then, with the standard multi-dimensional terminal data as input and the standard delivery records as output, a training data set is prepared, and a label mapping decision network is designed and trained. Exemplarily, the label mapping decision network is constructed through a deep learning model.

[0060] Furthermore, multiple audience label prediction channels are connected in parallel, and the input ends of the multiple audience label prediction channels are connected to the distributor to ensure that each channel independently predicts its corresponding delivery grid; then, the output of each audience label prediction channel is connected to the input end of the label mapping decision network to form a complete label mapping model, so that the label mapping model can predict the real-time label clusters of multiple delivery grids and the label clusters of corresponding required advertisements in real time, thereby improving the accuracy of the delivery strategy.

[0061] The advertising task processing module 15 is used to obtain the advertising task information of the target optimization block, extract the advertising task form and the task timestamp, and input the task timestamp into the label mapping model to obtain multiple advertising label clusters of multiple delivery grids.

[0062] Specifically, the current advertising task information of the target optimization block is obtained, and the advertising task form and task timestamp are determined. The advertising task form extracts relevant task information from the advertising task database, including advertising content, target audience, display duration, etc., and the task timestamp is the future time node that needs to be displayed.

[0063] Specifically, the task timestamp is input into the label mapping model for analysis, the advertising label clusters of different delivery grids in a specific time period are predicted, and the advertising label clusters of multiple grids are generated to reflect the audience characteristics and behavior patterns of each grid in the time period.

[0064] The delivery plan optimization module 16 is used to perform optimal matching between the advertisement task form and the multiple advertisement tag clusters based on a multi-objective optimization algorithm to obtain an advertisement delivery plan.

[0065] Specifically, a multi-objective optimization algorithm is used to achieve optimal matching between an advertising task form and multiple advertising tag clusters, wherein exemplary optimization objectives include maximizing advertising exposure or click-through rate, minimizing delivery costs, improving the matching degree of the target audience, and increasing the diversity of advertising delivery, etc.; from the solution set obtained during the optimization process, the solution that best meets business needs can be selected to achieve efficient advertising delivery.

[0066] In some embodiments, based on a multi-objective optimization algorithm, optimal matching is performed between the advertising task form and the plurality of advertising tag clusters to obtain an advertising delivery plan, and the execution steps include:

[0067] Establish task label clusters for multiple advertising tasks in the advertising task form; construct a scheme evaluation function, which includes a label matching factor, a richness factor and a task satisfaction factor; take multiple task label clusters as matching targets, traverse multiple advertising label clusters for non-repetitive random sampling matching, until all task label clusters are matched with corresponding advertising label clusters, and obtain alternative advertising delivery plans; iterate to obtain multiple alternative advertising delivery plans, and use the scheme evaluation function as the objective function to optimally select multiple alternative advertising delivery plans, and output them as the advertising delivery plan.

[0068] Specifically, first, extract task features from the advertising task form and establish a task label cluster for each task. The task label cluster is a collection of characteristic indicators of the advertisements to be delivered, such as the target audience of the task, content features, etc.; then, construct label matching factors, richness factors and task satisfaction factors respectively, and fuse them through a weighted method to form a scheme evaluation function, among which the label matching factor is used to measure the matching degree between the advertising label cluster and multiple labels of the task label cluster, and the richness factor is used to evaluate the diversity and coverage of advertising labels in each scheme; the task satisfaction factor is used to measure the satisfaction degree of the advertising task requirements (such as delivery duration, population coverage, etc.).

[0069] Exemplarily, the label matching factor is constructed based on the cosine similarity of the label matching results in the alternative advertising delivery plans, the richness factor is defined by calculating the entropy of the labels in the alternative advertising delivery plans, and the task satisfaction factor is determined by weighted summation of delivery duration, coverage and other indicators.

[0070] Specifically, a non-repetitive random sampling is performed on multiple advertising label clusters, and the task label clusters are traversed one by one to ensure that each task label cluster has a corresponding advertising label cluster. After completing the initial matching, a set of alternative advertising delivery plans are obtained; then, the above process is repeated iteratively to continuously generate new alternative plans, and the plan evaluation function is used to calculate the score of each alternative plan, and then the plan with the highest score is selected from multiple alternative plans as the optimal advertising delivery plan.

[0071] Through the above steps, we can fully consider various delivery needs, effectively achieve the optimal match between advertising tasks and label clusters, and optimize the advertising delivery effect.

[0072] In some embodiments, the execution steps of the system further include:

[0073] Based on a preset tracing cycle, multi-dimensional terminal feedback data is collected in real time, wherein the multi-dimensional terminal feedback data includes multi-dimensional personal terminal feedback data and multi-dimensional public terminal feedback data; the multi-dimensional personal terminal feedback data is parsed, and real-time user tag clusters of multiple delivery grids are extracted to obtain feedback tag data; the feedback tag data is compared with the audience tag prediction result to calculate the audience tag deviation; if the audience tag deviation is greater than or equal to the tag deviation control limit for a preset number of consecutive times, the corresponding multi-dimensional terminal feedback data is transmitted to the audience tag prediction channel for feedback optimization.

[0074] Specifically, the preset tracing period is the time period for regularly collecting feedback data, and the obtained multi-dimensional terminal feedback data includes multi-dimensional personal terminal feedback data and multi-dimensional public terminal feedback data, wherein the multi-dimensional personal terminal feedback data includes multi-dimensional information of multiple personal terminals at time nodes corresponding to the preset tracing period, wherein the specific frequency of feedback analysis and adjustment (i.e., the preset tracing period) is determined based on the delivery demand and platform computing power of the target Internet advertising platform. By way of example, for Internet advertising platforms with sufficient platform computing power and high delivery effect requirements, a higher analysis and adjustment frequency can be configured, such as minutes or seconds.

[0075] Specifically, the corresponding real-time user label cluster is extracted from the multi-dimensional personal terminal feedback data, and the real-time user label cluster represents the real-time audience characteristics of multiple delivery grids; then, the feedback label data is compared with the original audience label prediction results, and the audience label deviation is calculated to measure the difference between the prediction and the actual; then, according to the set label deviation control limit, whether the deviation exceeds the preset number of times continuously is monitored. If the deviation exceeds the control limit for more than the preset number of times, the corresponding multi-dimensional terminal feedback data is transmitted to the audience label prediction channel for optimization.

[0076] Optionally, the audience label deviation is calculated based on the difference in confidence values ​​of multiple label indicators, such as weighted averaging or summing the difference between the confidence values ​​of the feedback label data and all corresponding label indicators in the original audience label prediction results to obtain the overall audience label deviation.

[0077] Through the above steps, the system can dynamically adjust the prediction performance of the modified audience tag prediction channel, thereby ensuring higher advertising delivery accuracy and effectiveness.

[0078] In some embodiments, the execution steps of the system further include:

[0079] According to the audience tag deviation, a cumulative deviation rate is obtained through statistical analysis. If the cumulative deviation rate is greater than or equal to a preset cumulative deviation control limit, the corresponding plurality of multi-dimensional terminal feedback data are transmitted to the audience tag prediction channel for feedback optimization.

[0080] Specifically, a cumulative deviation analysis is performed based on the audience label deviation. For example, an accumulator is used to accumulate the number of times that the audience label deviation is greater than or equal to the label deviation control limit, and the proportion of the number of times is calculated accordingly, and the output is the cumulative deviation rate. If the cumulative deviation rate is greater than or equal to the preset cumulative deviation control limit (such as 5%), it can be considered that the prediction accuracy of the audience label prediction channel is poor. The same steps and ideas as above are used to transmit the corresponding multiple multi-dimensional terminal feedback data to the audience label prediction channel for feedback optimization.

[0081] In some embodiments, the execution steps of the system further include:

[0082] Parse the multi-dimensional terminal feedback data to extract delivery performance data; compare the delivery performance data with expected performance data based on the advertising task information, and if the delivery performance data does not meet the expected performance data, perform enhanced training of the label mapping decision network based on the multi-dimensional terminal feedback data.

[0083] Furthermore, delivery performance data is extracted from the multi-dimensional terminal feedback data, and the actual delivery performance data is compared with the expected performance data using the same principle as the above feedback label data and the original audience label prediction results, wherein the expected performance data is determined based on the advertising task information.

[0084] Specifically, if the delivery performance data does not meet the expected performance data, the multi-dimensional terminal feedback data is used to perform reinforcement training on the label mapping decision network. Exemplarily, the model parameters are updated through a reinforcement learning algorithm (such as Q-learning or deep reinforcement learning) to enhance the label mapping and decision-making capabilities of the label mapping decision network to ensure continuous optimization and improve the delivery effect.

[0085] Preferably, the delivery performance data extracted from the multi-dimensional terminal feedback data obtained after the advertisement is delivered can be used as the basis for generating and adjusting the advertisement task form. For example, according to the CTR (click-through rate), CVR (conversion rate), eCPM (effective revenue per thousand impressions), ROI (return on investment) and other indicators in the delivery performance data, the performance of different advertisement positions can be determined, and according to the performance differences of multiple advertisement positions, the bidding strategies corresponding to the multiple advertisement positions can be adjusted. For example, if the conversion rate in a certain region is high, the reserve price in that region will be increased. Extracting the delivery performance data through the multi-dimensional terminal feedback data can provide strong data support for the generation and adjustment of the advertisement task form, especially in the real-time bidding (RTB) environment, which helps advertisers to dynamically optimize bidding strategies, budget allocation and advertisement content, so as to realize the precision of advertisement delivery and maximize the benefits.

[0086] In summary, the AI ​​marketing advertising optimization system integrating multi-dimensional data provided by the present invention has the following technical effects:

[0087] Through the historical record acquisition module for determining the target optimization block and obtaining its historical delivery records; the terminal data interaction module for interacting with the advertising delivery management platform and obtaining relevant multi-dimensional terminal data according to the historical delivery records, including multi-dimensional personal terminal data and multi-dimensional public terminal data; the grid division module for gridding the target optimization block and obtaining grid time series label data; the mapping model construction module for building a label mapping model based on grid time series label data, historical delivery records and multi-dimensional terminal data, including multiple parallel audience label prediction channels and label mapping decision networks; the advertising task processing module for extracting advertising task information, task forms and timestamps, and inputting the timestamps into the label mapping model to obtain advertising label groups of multiple delivery grids; the delivery plan optimization module for matching advertising task forms and multiple advertising label groups through a multi-objective optimization algorithm to obtain the best advertising delivery plan. The technical effects of improving data utilization, providing predictive delivery optimization and increasing delivery revenue are achieved.

[0088] Example 2

[0089] Figure 2 This is a flowchart of an AI marketing advertising optimization method integrating multi-dimensional data in the present invention. For example, Figure 1 The structural diagram of the AI ​​marketing advertising optimization system integrating multi-dimensional data in the present invention can be used to perform the following steps: Figure 2 The process shown.

[0090] Based on the same concept as the AI ​​marketing advertising delivery optimization system integrating multi-dimensional data in the embodiment, the present invention also provides an AI marketing advertising delivery optimization method integrating multi-dimensional data, including:

[0091] Determine the target optimization block and obtain the historical delivery records of the target optimization block.

[0092] Based on the historical delivery records, the interactive advertisement delivery management platform obtains associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data.

[0093] The target optimization block is grid-divided to obtain a plurality of delivery grids, and a plurality of grid time series label clusters of the plurality of delivery grids are constructed to obtain grid time series label data.

[0094] A label mapping model is constructed based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, wherein the label mapping model includes a plurality of parallel audience label prediction channels and a label mapping decision network.

[0095] The advertisement task information of the target optimization block is obtained, the advertisement task form and the task timestamp are extracted, and the task timestamp is input into the label mapping model to obtain a plurality of advertisement label clusters of the plurality of delivery grids.

[0096] Based on a multi-objective optimization algorithm, the advertising task form is optimally matched with the multiple advertising tag clusters to obtain an advertising delivery plan.

[0097] In some embodiments, based on the historical delivery records, the interactive advertising delivery management platform obtains associated multi-dimensional terminal data, including:

[0098] Based on the target optimized block, the corresponding block position information and block window information are obtained; with the block position information and the block window information as call constraints, the advertising delivery management platform is interacted with to extract the audience effectiveness data and audience tag data of multiple personal terminals, and output as the multi-dimensional personal terminal data; based on the historical delivery records, the corresponding multiple public terminals are matched on the advertising delivery management platform, and the public effectiveness data of the multiple public terminals are extracted, and output as the multi-dimensional public terminal data.

[0099] In some embodiments, the target optimization block is gridded, a plurality of delivery grids are obtained, and a plurality of grid time series label clusters of the plurality of delivery grids are constructed to obtain grid time series label data, including:

[0100] The grid granularity of the target optimization block is defined, and based on the grid granularity, the target optimization block is divided into multiple delivery grids; multiple groups of multidimensional terminal grid data corresponding to the multiple delivery grids are obtained by matching the multidimensional terminal data; based on the sliding window method, multiple groups of the multidimensional terminal grid data are statistically analyzed, and multiple grid time series label clusters of the multiple delivery grids are extracted, and output as the grid time series label data.

[0101] In some embodiments, a label mapping model is constructed based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, including:

[0102] Statistically analyzing the periodic characteristics of multiple grid time series label clusters in the grid time series label data, and constructing multiple audience label prediction channels; defining a delivery efficiency constraint based on the expected delivery effect of the target optimization block, and performing data cleaning of the historical delivery records and the multidimensional terminal data according to the delivery efficiency constraint to obtain standard delivery records and standard multidimensional terminal data; using the standard multidimensional terminal data as training input data and the standard delivery records as training target output, constructing and training the label mapping decision network; connecting multiple audience label prediction channels in parallel, and generating the label mapping model by connecting the output ends of multiple audience label prediction channels with the input end of the label mapping decision network, wherein multiple parallel audience label prediction channels are used to respectively predict the real-time label clusters of multiple delivery grids.

[0103] In some embodiments, based on a multi-objective optimization algorithm, optimal matching of the advertisement task form with the plurality of advertisement tag clusters is performed to obtain an advertisement delivery plan, including:

[0104] Establish task label clusters for multiple advertising tasks in the advertising task form; construct a scheme evaluation function, which includes a label matching factor, a richness factor and a task satisfaction factor; take multiple task label clusters as matching targets, traverse multiple advertising label clusters for non-repetitive random sampling matching, until all task label clusters are matched with corresponding advertising label clusters, and obtain alternative advertising delivery plans; iterate to obtain multiple alternative advertising delivery plans, and use the scheme evaluation function as the objective function to optimally select multiple alternative advertising delivery plans, and output them as the advertising delivery plan.

[0105] In some embodiments, the method further comprises:

[0106] Based on a preset tracing cycle, multi-dimensional terminal feedback data is collected in real time, wherein the multi-dimensional terminal feedback data includes multi-dimensional personal terminal feedback data and multi-dimensional public terminal feedback data; the multi-dimensional personal terminal feedback data is parsed, and real-time user tag clusters of multiple delivery grids are extracted to obtain feedback tag data; the feedback tag data is compared with the audience tag prediction result to calculate the audience tag deviation; if the audience tag deviation is greater than or equal to the tag deviation control limit for a preset number of consecutive times, the corresponding multi-dimensional terminal feedback data is transmitted to the audience tag prediction channel for feedback optimization.

[0107] In some embodiments, the method further comprises:

[0108] Parse the multi-dimensional terminal feedback data to extract delivery performance data; compare the delivery performance data with expected performance data based on the advertising task information, and if the delivery performance data does not meet the expected performance data, perform enhanced training of the label mapping decision network based on the multi-dimensional terminal feedback data.

[0109] Example 3

[0110] The present embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: determining a target optimization block, and obtaining a historical delivery record of the target optimization block. Based on the historical delivery record, the interactive advertising delivery management platform obtains associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data. The target optimization block is grid-divided, a plurality of delivery grids are obtained, and a plurality of grid time-series label clusters of the plurality of delivery grids are constructed to obtain grid time-series label data. Based on the grid time-series label data, the historical delivery record and the multi-dimensional terminal data, a label mapping model is constructed, wherein the label mapping model includes a plurality of parallel audience label prediction channels and a label mapping decision network. The advertising task information of the target optimization block is obtained, the advertising task form and the task timestamp are extracted, and the task timestamp is input into the label mapping model to obtain a plurality of advertising label clusters of the plurality of delivery grids. Based on the multi-objective optimization algorithm, the advertising task form is optimally matched with the plurality of advertising label clusters to obtain an advertising delivery plan.

[0111] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the AI ​​marketing advertising optimization method integrating multi-dimensional data described in embodiment two. For the sake of brevity of the specification, they will not be further elaborated here.

[0112] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An AI marketing advertising optimization system integrating multi-dimensional data, characterized in that: The system comprises: A historical record acquisition module, which is used to determine the target optimization block and acquire the historical delivery record of the target optimization block; A terminal data interaction module, the terminal data interaction module is used to interact with the advertising delivery management platform based on the historical delivery records to obtain associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data; A grid division module, the grid division module is used to grid the target optimization block, obtain multiple delivery grids, and construct multiple grid time series label clusters of the multiple delivery grids to obtain grid time series label data; A mapping model construction module, the mapping model construction module is used to construct a label mapping model based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, wherein the label mapping model includes a plurality of parallel audience label prediction channels and a label mapping decision network; An advertising task processing module, the advertising task processing module is used to obtain advertising task information of a target optimization block, extract an advertising task form and a task timestamp, and input the task timestamp into the label mapping model to obtain multiple advertising label clusters of multiple delivery grids; The delivery plan optimization module is used to perform optimal matching between the advertisement task form and the multiple advertisement tag clusters based on a multi-objective optimization algorithm to obtain an advertisement delivery plan.

2. According to claim 1, an AI marketing advertising optimization system integrating multi-dimensional data is characterized in that: Based on the historical delivery records, the interactive advertising delivery management platform obtains the associated multi-dimensional terminal data, and the execution steps include: Based on the target optimized block, obtain the corresponding block position information and block window information; Taking the block position information and the block window information as call constraints, interacting with the advertising delivery management platform to extract audience effectiveness data and audience tag data of multiple personal terminals, and outputting them as the multi-dimensional personal terminal data; Based on the historical delivery records, the corresponding multiple public terminals are matched on the advertisement delivery management platform, and the public performance data of the multiple public terminals are extracted and output as the multi-dimensional public terminal data.

3. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 2 is characterized in that: The target optimization block is grid-divided, a plurality of delivery grids are obtained, and a plurality of grid time series label clusters of the plurality of delivery grids are constructed to obtain grid time series label data. The execution steps include: Defining a grid granularity of a target optimization block, and dividing the target optimization block into a plurality of delivery grids based on the grid granularity; Matching the multi-dimensional terminal data to obtain multiple groups of multi-dimensional terminal grid data corresponding to the multiple delivery grids; Based on the sliding window method, a plurality of groups of the multi-dimensional terminal grid data are statistically analyzed, and a plurality of the grid time series label clusters of a plurality of the delivery grids are extracted and output as the grid time series label data.

4. The AI ​​marketing advertising optimization system integrating multi-dimensional data according to claim 3 is characterized in that: Based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, a label mapping model is constructed, and the execution steps include: Statistically analyzing the periodic characteristics of the plurality of grid time series label clusters in the grid time series label data, and constructing a plurality of audience label prediction channels; Based on the expected delivery effect of the target optimization block, a delivery efficiency constraint is defined, and data cleaning of the historical delivery records and the multi-dimensional terminal data is performed according to the delivery efficiency constraint to obtain standard delivery records and standard multi-dimensional terminal data; Using the standard multi-dimensional terminal data as training input data and the standard delivery record as training target output, constructing and training the label mapping decision network; A plurality of the audience label prediction channels are connected in parallel, and the label mapping model is generated by connecting the output ends of the plurality of the audience label prediction channels with the input end of the label mapping decision network, wherein the plurality of parallel audience label prediction channels are used to respectively predict the real-time label clusters of a plurality of delivery grids.

5. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 4 is characterized in that: Based on a multi-objective optimization algorithm, optimal matching is performed between the advertising task form and the plurality of advertising tag clusters to obtain an advertising delivery plan. The execution steps include: Establishing a task tag cluster of multiple advertising tasks in the advertising task form; Constructing a scheme evaluation function, wherein the scheme evaluation function includes a label matching factor, a richness factor, and a task satisfaction factor; Taking the multiple task tag clusters as matching targets, traversing the multiple advertising tag clusters to perform non-repetitive random sampling matching until all task tag clusters are matched with corresponding advertising tag clusters, and obtaining alternative advertising delivery plans; Iteratively obtain a plurality of the candidate advertising delivery plans, and use the plan evaluation function as the objective function to optimally select the plurality of the candidate advertising delivery plans, and output the advertising delivery plan.

6. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 1 is characterized in that: The execution steps of the system also include: Based on a preset tracing cycle, multi-dimensional terminal feedback data is collected in real time, wherein the multi-dimensional terminal feedback data includes multi-dimensional personal terminal feedback data and multi-dimensional public terminal feedback data; Parsing the multi-dimensional personal terminal feedback data, extracting real-time user tag clusters of a plurality of the delivery grids, and obtaining feedback tag data; The feedback label data is compared with the audience label prediction result to calculate the audience label deviation.

7. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 6 is characterized in that: The execution steps of the system also include: If the audience label deviation is greater than or equal to the label deviation control limit for a preset number of consecutive times, the corresponding multi-dimensional terminal feedback data is transmitted to the audience label prediction channel for feedback optimization.

8. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 6 is characterized in that: The execution steps of the system also include: According to the audience tag deviation, a cumulative deviation rate is obtained through statistical analysis. If the cumulative deviation rate is greater than or equal to a preset cumulative deviation control limit, the corresponding plurality of multi-dimensional terminal feedback data are transmitted to the audience tag prediction channel for feedback optimization.

9. The AI ​​marketing advertising delivery optimization system integrating multi-dimensional data according to claim 8 is characterized in that: The execution steps of the system also include: Analyzing the multi-dimensional terminal feedback data to extract delivery effectiveness data; The delivery performance data is compared with the expected performance data based on the advertising task information. If the delivery performance data does not meet the expected performance data, the label mapping decision network is subjected to enhanced training based on the multi-dimensional terminal feedback data.

10. An AI marketing advertising optimization method integrating multi-dimensional data, characterized in that: The method is applied to an AI marketing advertising delivery optimization system integrating multi-dimensional data as described in any one of claims 1 to 9, and the method comprises: Determine the target optimization block and obtain the historical delivery records of the target optimization block; Based on the historical delivery records, the interactive advertising delivery management platform obtains associated multi-dimensional terminal data, wherein the multi-dimensional terminal data includes multi-dimensional personal terminal data and multi-dimensional public terminal data; Gridding the target optimization block, obtaining a plurality of delivery grids, and constructing a plurality of grid time series label clusters of the plurality of delivery grids to obtain grid time series label data; Based on the grid time series label data, the historical delivery records and the multi-dimensional terminal data, a label mapping model is constructed, wherein the label mapping model includes a plurality of parallel audience label prediction channels and a label mapping decision network; Acquire the advertising task information of the target optimization block, extract the advertising task form and task timestamp, and input the task timestamp into the label mapping model to obtain multiple advertising label clusters of multiple delivery grids; Based on a multi-objective optimization algorithm, the advertising task form is optimally matched with the multiple advertising tag clusters to obtain an advertising delivery plan.

Images