Cross-platform advertisement effect real-time digital tracking and analysis system

Abstract

The invention provides a cross-platform advertisement effect real-time digital tracking and analysis system, and relates to the technical field of Internet information, and the system comprises a time sequence modeling module of cross-platform user behaviors, a user behavior data collection module under differential privacy protection, and a cross-platform advertisement effect correlation analysis module based on a graph neural network. Through a time sequence behavior model, the system can predict future advertisement interaction behaviors of a user on different platforms, so that an advertisement putting strategy is optimized, the safety of user behavior data is ensured through a differential privacy technology, an advertiser or a third party is prevented from inferring specific user information from an advertisement effect, the privacy protection requirement is met, and the user experience is improved. Through cross-platform analysis of the graph neural network, the system can discover the synergistic effect of the advertisement on different platforms, intelligently distribute advertisement resources and improve the overall putting effect and conversion rate of the advertisement, and through a real-time feedback mechanism, an advertiser can optimize an advertisement strategy according to data provided by the system and realize dynamic adjustment of advertisement putting.

Description

Technical Field

[0001] The present invention relates to the field of Internet information technology, and in particular to a real-time digital tracking and analysis system for cross-platform advertising effects. Background Art

[0002] According to a network advertising effect evaluation system and method based on eye tracking disclosed in Chinese patent number "CN106920129B", the system includes: an eye tracking collection unit, which is used to obtain user test behavior information on a search engine result page through an eye tracking device; an advertising effect model building unit, which is used to qualitatively and quantitatively analyze the effects of advertisements with different positions and relevances on user attention and attractiveness based on the data obtained by the eye tracking collection unit; and based on the results of the eye movement analysis, combined with the characteristics of the advertisement itself, displayed feedback information and implicit feedback information, extract features related to the network advertising effect to obtain an advertising effect model; an advertising effect evaluation unit, which is used to calculate the relationship between the advertising effect and each feature based on the advertising effect model, and evaluate the effectiveness of the advertising effect model.

[0003] The above patent documents and prior art have the following technical problems when used:

[0004] Question 1: The above patent documents and existing technologies mainly rely on eye tracking technology to obtain indicators such as the attractiveness and attention of advertisements by analyzing the test behavior of users on the search engine results page. However, this method relies too much on the visual information of users and ignores the impact of users on other platforms and based on various interactive behaviors, such as clicks, browsing, and purchases, on the advertising effect. Although eye tracking can reflect the short-term visual behavior of users on a single platform, it is difficult to capture the global effect of cross-platform advertising.

[0005] Problem 2: When capturing user behavior data in the above documents, the user's visual data and behavior path records are directly involved, which will lead to the risk of privacy leakage. There is a lack of privacy protection mechanism, and the risk of user behavior data being exposed is high;

[0006] Question three: In the above-mentioned patent documents and prior arts, when tracking highlights, they are limited to a single platform, the search engine results page, and do not take into account users’ advertising interaction behaviors on different platforms and cross-platform advertising synergy effects. Summary of the invention

[0007] Technical issues solved

[0008] In view of the shortcomings of the prior art, the present invention provides a real-time digital tracking and analysis system for cross-platform advertising effects, which solves the following problems:

[0009] 1. The user behavior data dimension is single and cannot reflect the overall advertising effect;

[0010] 2. Regarding the issue that user privacy is not adequately protected;

[0011] 3. Address the problem of lack of coordinated optimization of cross-platform advertising effects.

[0012] Technical Solution

[0013] To achieve the above objectives, the present invention is implemented through the following technical solutions: a real-time digital tracking and analysis system for cross-platform advertising effects, the system includes a time series modeling module for cross-platform user behavior, a user behavior data collection module under differential privacy protection, and a cross-platform advertising effect association analysis module based on graph neural network, wherein:

[0014] The cross-platform user behavior time series modeling module models the collected cross-platform data through a hidden Markov model. The system constructs a time series of the user's advertising interaction behavior and advertising behavior status at different times and on different platforms, and predicts the user's future advertising behavior in real time based on past behavior to capture the user's behavior change trend;

[0015] The user behavior data collection module under differential privacy protection collects users' advertising interaction data from various platforms through differential privacy technology, so that the system adds random noise when querying user behavior, ensuring that the user's behavior cannot be specifically identified;

[0016] The cross-platform advertising effect correlation analysis module based on graph neural network, the system will regard each platform as a node, the user's cross-platform behavior as the edge between the nodes, establish a cross-platform advertising effect correlation model through the graph neural network, and analyze the correlation of users' advertising interaction behaviors on different platforms;

[0017] The main control module receives and analyzes the data of each module, regularly updates and evaluates the advertising effect, and feeds back the evaluation structure to each module for training.

[0018] Preferably, in the cross-platform user behavior temporal modeling module, the user's advertising interaction behavior state at time t is set as S t , through the hidden Markov model, the transition probability of user behavior is:

[0019]

[0020] in:

[0021] P(S t |S t-1 , S t-2 ,…,S1) represents the probability of the user’s advertising behavior state at time t under the condition of a given historical state;

[0022] S t The state of advertising behavior at time t;

[0023] P(S t , S t-1 ,…,S1) represents the joint probability, which indicates the probability of occurrence of the behavior sequence from time 1 to t;

[0024] P(S t-1 ,…,S1) represents the total probability of the user’s behavior sequence from time 1 to t-1.

[0025] Preferably, in the user behavior data collection module under differential privacy protection, noise N(0, σ 2 ), namely the differential privacy mechanism:

[0026] M(D)=f(D)+N(0,σ 2 )

[0027] in:

[0028] N(0,σ 2 ) represents Gaussian noise;

[0029] M(D) represents the output under the differential privacy mechanism;

[0030] D represents the user behavior dataset;

[0031] f(D) represents the query or statistical function of the data set;

[0032] N(0,σ 2 ) means the mean is 0 and the variance is σ 2 Gaussian noise.

[0033] Preferably, the cross-platform advertising effect correlation analysis module based on graph neural network regards the advertising on different platforms as a graph structure, with the effect of advertising displayed on each platform as a node, the correlation between platforms as an edge, and the advertising effect H i The update on the i-th node is expressed as:

[0034]

[0035] in:

[0036] represents the advertising effect status of the ith base point at time t;

[0037] N(i) represents the set of neighbor nodes connected to the i-th node;

[0038] W ij represents the strength of association between platform i and platform j;

[0039] f(*) represents the activation function for updating the node state.

[0040] Preferably, the cross-platform user behavior temporal modeling module analyzes the user's interactive behavior towards advertisements on multiple platforms, and the interactive behavior towards advertisements includes the user's single click behavior, long-term behavior patterns, interest changes, and behavioral associations across platforms, and the advertising behavior status includes clicks, browsing, and purchases.

[0041] Preferably, the main control module receives feedback information from the cross-platform user behavior time series modeling module and the cross-platform advertising effect association analysis module based on graph neural network, and the feedback information includes the user's future advertising interaction probability and Guangao click-through rate on a certain platform. The main control module updates the data every day or every week, inputs the new user behavior data into the hidden Markov model and the graph neural network model, and recalculates the advertising effect prediction and cross-platform correlation.

[0042] Preferably, the operation process of the system is as follows:

[0043] Sp1: Data collection and preprocessing. The user behavior data collection module under differential privacy protection collects user behavior data and adds differential privacy protection to the user behavior data during collection;

[0044] Sp2: Cross-platform temporal behavior modeling. The cross-platform user behavior temporal modeling module receives the encrypted user behavior data of the Sp1 mobile phone, and constructs a time series of the user's behavior at different times and on different platforms through hidden Markov modeling to capture the user's behavior change trend;

[0045] Sp3: Cross-platform association analysis. The cross-platform advertising effect association analysis module based on graph neural network regards each platform as a node, and the user's cross-platform behavior as the edge between nodes. The association model of cross-platform advertising effect is established through the graph neural network, and the association of user's advertising interaction behavior on different platforms is analyzed through the graph neural network.

[0046] Sp4: Real-time feedback and advertising strategy optimization. The main control module receives the associated data of each module and provides feedback and optimization. The system dynamically adjusts the display frequency of advertisements based on user behavior predictions. Based on the results of graph neural network analysis, the system can discover the synergy of cross-platform advertisements.

[0047] Sp5: Continuous learning and model updating. The system updates data every day or every week, inputs new user behavior data into the time series model and graph neural network, recalculates advertising effect predictions and cross-platform correlations, and uses the latest data to dynamically adjust model parameters to ensure that the system can adapt to changing user behaviors and market dynamics.

[0048] Preferably, the types of data collected in the data collection and preprocessing include user behavior data, platform data and timestamp data, and the data sources include client data, server data and user devices.

[0049] Beneficial Effects

[0050] The present invention provides a real-time digital tracking and analysis system for cross-platform advertising effects. It has the following beneficial effects:

[0051] 1. The present invention adopts a system that uses hidden Markov model technology to perform time series modeling on users' advertising interactive behaviors on various platforms, capture users' behavioral patterns at different time points, and achieve accurate prediction of advertising effects. It can estimate users' future advertising clicks, browsing or purchasing behaviors, thereby helping advertisers optimize advertising delivery time and frequency according to users' behavioral habits and improve the conversion rate and effect of advertising delivery.

[0052] 2. The present invention uses differential privacy protection technology in the system to add noise to user behavior data, ensuring that individual user data cannot be identified or tracked, preventing advertisers or third parties from obtaining specific user information from advertising effect data, thereby achieving effective protection of user privacy and ensuring that user personal information will not be leaked during advertising effect analysis and behavior tracking, meeting the needs of data security and privacy protection, and enhancing user trust in advertising.

[0053] 3. The present invention uses graph neural network technology within the system to construct a graph structure of cross-platform user behavior, associate the advertising effects on different platforms through graph nodes and edges, and analyze the cross-platform correlation of user behavior through the message propagation mechanism of GNN, thereby realizing the collaborative optimization of cross-platform advertising. Advertisers can discover the synergy between platforms based on the interactive behavior of users on different platforms, optimize the allocation of advertising resources, and maximize the synergy of advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a system structure diagram of the present invention;

[0055] Figure 2 It is a flow chart of the system operation of the present invention;

[0056] Figure 3 A visualization chart of a hidden Markov model of the temporal modeling module of the cross-platform user behavior of the present invention;

[0057] Figure 4 A differential privacy visualization chart of the user behavior data collection module under differential privacy protection of the present invention;

[0058] Figure 5 The cross-platform advertising effect association analysis module based on graph neural network of the present invention

[0059] Figure 6 A visual chart showing the advertisement effect update and feedback of the main control module of the present invention;

[0060] Figure 7 A visual line chart of the cross-platform advertising effect of the system of the present invention;

[0061] Figure 8 A visualized discrete graph of the cross-platform advertising effect of the system of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Specific embodiment one:

[0064] like Figure 1-8 As shown, a real-time digital tracking and analysis system for cross-platform advertising effects includes a time series modeling module for cross-platform user behavior, a user behavior data collection module under differential privacy protection, and a cross-platform advertising effect correlation analysis module based on graph neural network, wherein:

[0065] The cross-platform user behavior time series modeling module models the collected cross-platform data through the hidden Markov model. The system forms a time series of users' advertising interaction behaviors and advertising behavior states at different times and on different platforms, and predicts users' future advertising behaviors in real time based on past behaviors to capture the changing trends of users' behavior. In the cross-platform user behavior time series modeling module, the user's advertising interaction behavior state at time t is set as S t , through the hidden Markov model, the transition probability of user behavior is:

[0066]

[0067] in:

[0068] P(S t |S t-1 , S t-2 ,…,S1) represents the probability of the user’s advertising behavior state at time t under the condition of a given historical state;

[0069] S t The state of advertising behavior at time t;

[0070] P(S t , S t-1 ,…,S1) represents the joint probability, which indicates the probability of occurrence of the behavior sequence from time 1 to t;

[0071] P(S t-1 ,…,S1) represents the total probability of the user's behavior sequence from time 1 to t-1;

[0072] The user behavior data collection module under differential privacy protection collects users’ advertising interaction data from various platforms through differential privacy technology, so that the system adds random noise when querying user behavior to ensure that the user’s behavior will not be specifically identified. In the user behavior data collection module under differential privacy protection, noise N(0, σ 2 ), namely the differential privacy mechanism:

[0073] M(D)=f(D)+N(0,σ 2 )

[0074] in:

[0075] N(0,σ 2 ) represents Gaussian noise;

[0076] M(D) represents the output under the differential privacy mechanism;

[0077] D represents the user behavior dataset;

[0078] f(D) represents the query or statistical function of the data set;

[0079] N(0,σ 2 ) means the mean is 0 and the variance is σ 2 Gaussian noise;

[0080] The cross-platform advertising effect correlation analysis module based on graph neural network regards each platform as a node, and the user's cross-platform behavior as the edge between nodes. Through the graph neural network, a correlation model of cross-platform advertising effects is established to analyze the correlation of user advertising interaction behaviors on different platforms. The cross-platform advertising effect correlation analysis module based on graph neural network regards the advertising placement on different platforms as a graph structure, with the effect of advertising display on each platform as a node, and the correlation between platforms as an edge. i The update on the i-th node is expressed as:

[0081]

[0082] in:

[0083] represents the advertising effect status of the ith base point at time t;

[0084] N(i) represents the set of neighbor nodes connected to the i-th node;

[0085] W ij represents the strength of association between platform i and platform j;

[0086] f(*) represents the activation function for updating the node state;

[0087] The main control module receives and analyzes the data from each module, regularly updates and evaluates the advertising effect, and feeds the evaluation structure back to each module for training.

[0088] Through the temporal behavior model, the entire system mentioned above can predict the user's future advertising interaction behavior on different platforms, thereby optimizing the advertising delivery strategy. Differential privacy technology ensures the security of user behavior data, prevents advertisers or third parties from inferring specific user information from advertising effects, and meets privacy protection requirements. Through cross-platform analysis of graph neural networks, the system discovers the synergy of advertising on different platforms, intelligently allocates advertising resources, and improves the overall advertising delivery effect and conversion rate. Through the real-time feedback mechanism, advertisers can optimize advertising strategies based on the data provided by the system and realize dynamic adjustment of advertising delivery. It integrates a number of advanced technologies and can achieve efficient and privacy-protected real-time tracking and analysis of advertising effects, which has a significant optimization effect on advertisers' marketing strategies and user experience. Specific embodiment 2:

[0090] like Figure 1-8 As shown, according to the contents in the above specific embodiments, the following contents are further disclosed:

[0091] The cross-platform user behavior temporal modeling module analyzes the user's interactive behavior towards advertisements on multiple platforms. The interactive behavior towards advertisements includes the user's single click behavior, long-term behavior patterns, interest changes, and behavioral associations across platforms. The advertising behavior status includes clicks, browsing, and purchases. By combining the user's behavioral data on different platforms, a hidden Markov model is trained to predict the user's future advertising behavior in real time. Through such temporal analysis, the user's cross-platform behavior patterns can be effectively captured, and their possible reactions in subsequent advertising interactions can be predicted.

[0092] In the user behavior data collection module under differential privacy protection, the core idea of ​​differential privacy is: when two adjacent data sets, that is, only one record different, perform the same query, the difference in the output results will not be significant, so specific user information cannot be inferred from the query results. The differential privacy mechanism is achieved by adding noise. For each query of user behavior data such as click-through rate and viewing time, Gaussian noise is added. Federated learning technology is used to process data through multiple distributed nodes to avoid uploading raw data to the central server. This module ensures that the advertising system will not leak users' personal privacy when tracking across platforms. The introduction of differential privacy technology enables advertisers to meet the requirements of privacy regulations such as GDPR when obtaining valuable advertising effect data, thereby enhancing user trust and the legality and compliance of the platform.

[0093] The cross-platform advertising effect correlation analysis module based on graph neural network establishes a graph structure model for advertising on different platforms, regards each platform as a node, and the user's cross-platform behavior as an edge, uses GNN to update the node status, and analyzes the interaction and correlation of advertising effects between platforms. Through graph neural network analysis, advertisers can understand the synergistic effect of advertising on different platforms. For example, when an advertisement performs well on social media, it may affect users' purchasing behavior on e-commerce platforms. This module can optimize the cross-platform resource allocation of advertising and improve the overall ROI of advertising.

[0094] The main control module receives feedback from the cross-platform user behavior time series modeling module and the cross-platform advertising effect correlation analysis module based on graph neural network. The feedback information includes the user's future advertising interaction probability and Guangao click-through rate on a certain platform. The main control module updates the data every day or every week, inputs the new user behavior data into the hidden Markov model and graph neural network model, and recalculates the advertising effect prediction and cross-platform correlation. Specific embodiment three:

[0096] like Figure 1-8 As shown, according to the contents in the above specific embodiments, the following contents are further disclosed:

[0097] The system operation process is as follows:

[0098] Sp1: Data collection and preprocessing. The user behavior data collection module under differential privacy protection collects user behavior data. The data includes users' advertising interaction behavior data on different platforms, such as social media, search engines, e-commerce platforms, etc., including click-through rate, browsing time, interaction frequency, conversion rate, etc., and adds differential privacy protection to user behavior data during collection. Since data from different platforms may have different formats or measurement units, it is necessary to unify the data structure of each platform through standardization processing, such as normalization and feature extraction. When querying data such as advertising click-through rate and browsing time, it is prevented that advertisers or platforms can infer specific user information from it, so as to meet privacy protection requirements;

[0099] Sp2: Cross-platform temporal behavior modeling. The cross-platform user behavior temporal modeling module receives the encrypted user behavior data from the Sp1 mobile phone, such as the time point of clicking on the advertisement, the advertisement viewing time, the click frequency, etc., as the time series input. Through the hidden Markov model modeling, the user's behavior at different times and on different platforms is formed into a time series to capture the user's behavior change trend. The system can predict the user's ad click behavior or purchase conversion behavior in the future time period by training on historical data.

[0100] Sp3: Cross-platform association analysis. The cross-platform advertising effect association analysis module based on graph neural network regards each platform as a node and the user's cross-platform behavior as the edge between nodes. The cross-platform advertising effect association model is established through the graph neural network. The association of user's advertising interaction behavior on different platforms is analyzed through the graph neural network. For example, after seeing an advertisement on a social platform, is it more likely for the user to make a purchase on an e-commerce platform? Through such analysis, advertisers can optimize their cross-platform advertising delivery strategies.

[0101] Sp4: Real-time feedback and advertising strategy optimization. The main control module receives the associated data of each module and provides feedback and optimization. The system dynamically adjusts the display frequency of advertisements based on user behavior predictions. Advertisers can adjust advertising strategies in real time based on these feedback information, such as increasing advertising exposure on a certain platform or reducing advertising on other platforms. Based on the results of graph neural network analysis, the system can discover the synergy of cross-platform advertising. For example, if the conversion effect of the advertisement seen by the user on the video platform is significant, the system can recommend increasing advertising on that platform.

[0102] Sp5: Continuous learning and model updating. The system updates data every day or every week, inputs new user behavior data into the time series model and graph neural network, recalculates advertising effect predictions and cross-platform correlations, and uses the latest data to dynamically adjust model parameters to ensure that the system can adapt to changing user behaviors and market dynamics. Specific embodiment four:

[0104] like Figure 1-8 As shown, according to the contents in the above specific embodiments, the following contents are further disclosed:

[0105] The data types collected during data collection and preprocessing include user behavior data, platform data, and timestamp data, and the data sources include client data, server data, and user devices, among which:

[0106] User behavior data: including the number of ad clicks, ad display duration, user stay time, page scroll depth, post-click behavior path, etc.;

[0107] Platform data: including user interaction logs, ad positions, display frequency and other information of each platform, which is used for environmental analysis of advertising effects;

[0108] Timestamp data: Each user behavior data will record the specific time to support time series analysis.

[0109] Client data collection: collect user behavior data on advertisements in real time through client SDKs embedded in various platforms;

[0110] Server-side data collection: Batch collect ad display, click and conversion data from the ad delivery platforms of various platforms;

[0111] User device: includes the user's device ID, browser type, operating system, etc., which is used for cross-device behavior tracking, but still needs to be anonymized to ensure privacy.

[0112] The cross-platform advertising effect association analysis module based on graph neural network uses graph neural network GNN modeling. GNN is a neural network model specially used to process graph structure data. In the cross-platform advertising effect analysis, we regard each platform as a node and the user's advertising interaction behavior between different platforms as edges to form a graph structure. GNN realizes the collaborative analysis and inference of cross-platform advertising effects through the propagation of information in the graph structure. The graph is constructed as follows:

[0113] Node: Each advertising platform is a node. Node characteristics include advertising performance data on the platform, such as click-through rate and interaction time;

[0114] Edge: The advertising interactions between users on different platforms form edges. The weight of the edge is determined by the user's interaction frequency on multiple platforms, path analysis, etc.

[0115] Through graph neural networks, we can analyze the dissemination effects of advertisements on different platforms. For example, after users see advertisements on social media, will they increase the probability of purchasing on e-commerce platforms? The system can identify platforms with better cross-platform advertising synergy and optimize resource allocation for advertising. Specific embodiment five:

[0117] like Figure 1-8 As shown, according to the contents in the above specific embodiments, the following contents are further disclosed:

[0118] After predicting user behavior and analyzing cross-platform advertising effects, the system will immediately feed back the results to advertisers to help them optimize their delivery strategies. The feedback includes predictions of users' future advertising interaction behaviors, evaluation of advertising synergies and effects on different platforms, and recommendations on advertising resource allocation between platforms. By comprehensively analyzing users' advertising behaviors on different platforms, the system can evaluate the performance of advertising on each platform as a whole and discover potential synergies between different platforms, thus achieving global optimization of advertising effects. Advertisers can formulate cross-platform advertising strategies from a macro perspective, take advantage of the advantages of multiple platforms, and achieve global maximization of advertising effects.

[0119] Ad frequency adjustment: Dynamically adjust the frequency of ad display based on the user's ad interaction prediction. For users who have frequently interacted with ads recently, the system may reduce ad display to avoid "ad fatigue";

[0120] Ad content adjustment: Based on the user behavior characteristics on different platforms, the system recommends that advertisers design differentiated advertising content for different platforms to improve conversion rates. The system automatically generates an advertising effectiveness analysis report, including the effectiveness of each platform, user behavior trend forecasts, and cross-platform synergy effects, etc., to help advertisers quickly obtain analysis results and simplify the advertising management process. Advertisers do not need to manually analyze the data of each platform. The system automatically completes the analysis and generates optimization suggestions, which significantly improves the efficiency of advertising management and the speed of decision-making.

[0121] The system will continuously update the parameters of the time series model and graph neural network (GNN) based on real-time user behavior data to adapt to changes in user behavior. The model will be retrained regularly, such as daily and weekly, to ensure that the model can promptly reflect the latest dynamics of user behavior. The system uses online learning methods to update model parameters in real time to reduce training delays. In order to prevent frequent model updates from consuming a large amount of system computing resources, incremental learning technology is used to update only some model parameters to reduce computing costs.

[0122] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprising a reference structure" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0123] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Real-time digital tracking and analysis system for cross-platform advertising effects, featuring: The system includes a cross-platform user behavior temporal modeling module, a user behavior data collection module under differential privacy protection, and a cross-platform advertising effect correlation analysis module based on graph neural network, wherein: The cross-platform user behavior time series modeling module models the collected cross-platform data through a hidden Markov model. The system constructs a time series of the user's advertising interaction behavior and advertising behavior status at different times and on different platforms, and predicts the user's future advertising behavior in real time based on past behavior to capture the user's behavior change trend; The user behavior data collection module under differential privacy protection collects users' advertising interaction data from various platforms through differential privacy technology, so that the system adds random noise when querying user behavior, ensuring that the user's behavior cannot be specifically identified; The cross-platform advertising effect correlation analysis module based on graph neural network, the system will regard each platform as a node, the user's cross-platform behavior as the edge between the nodes, establish a cross-platform advertising effect correlation model through the graph neural network, and analyze the correlation of users' advertising interaction behaviors on different platforms; The main control module receives and analyzes the data of each module, regularly updates and evaluates the advertising effect, and feeds back the evaluation structure to each module for training.

2. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: In the cross-platform user behavior temporal modeling module, the user's advertising interaction behavior state at time t is set as S t , through the hidden Markov model, the transition probability of user behavior is: in: P(S t |S t-1 , S t-2 ,…,S1) represents the probability of the user’s advertising behavior state at time t under the condition of a given historical state; S t The state of advertising behavior at time t; P(S t , S t-1 ,…,S1) represents the joint probability, which indicates the probability of occurrence of the behavior sequence from time 1 to t; P(S t-1 ,…,S1) represents the total probability of the user’s behavior sequence from time 1 to t-1.

3. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: In the user behavior data collection module under differential privacy protection, noise N(0, σ 2 ), namely the differential privacy mechanism: M(D)=f(D)+N(0,σ 2 ) in: N(0,σ 2 ) represents Gaussian noise; M(D) represents the output under the differential privacy mechanism; D represents the user behavior dataset; f(D) represents the query or statistical function of the data set; N(0,σ 2 ) means the mean is 0 and the variance is σ 2 Gaussian noise.

4. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: The cross-platform advertising effect correlation analysis module based on graph neural network regards the advertising on different platforms as a graph structure, with the effect of advertising displayed on each platform as a node, the correlation between platforms as an edge, and the advertising effect H i The update on the i-th node is expressed as: in: represents the advertising effect status of the ith base point at time t; N(i) represents the set of neighbor nodes connected to the i-th node; W ij represents the strength of association between platform i and platform j; f(*) represents the activation function for updating the node state.

5. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: The cross-platform user behavior temporal modeling module analyzes the user's interactive behavior towards advertisements on multiple platforms, and the interactive behavior towards advertisements includes the user's single click behavior, long-term behavior patterns, interest changes, and behavioral associations across platforms, and the advertisement behavior status includes clicks, browsing, and purchases.

6. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: The main control module receives feedback information from the cross-platform user behavior time series modeling module and the cross-platform advertising effect association analysis module based on graph neural network, and the feedback information includes the user's future advertising interaction probability and advertising click-through rate on a certain platform. The main control module updates data every day or every week, inputs new user behavior data into the hidden Markov model and the graph neural network model, and recalculates the advertising effect prediction and cross-platform correlation.

7. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 1, characterized in that: The operation process of the system is as follows: Sp1: Data collection and preprocessing. The user behavior data collection module under differential privacy protection collects user behavior data and adds differential privacy protection to the user behavior data during collection; Sp2: Cross-platform temporal behavior modeling. The cross-platform user behavior temporal modeling module receives the encrypted user behavior data of the Sp1 mobile phone, and constructs a time series of the user's behavior at different times and on different platforms through hidden Markov modeling to capture the user's behavior change trend; Sp3: Cross-platform association analysis. The cross-platform advertising effect association analysis module based on graph neural network regards each platform as a node, and the user's cross-platform behavior as the edge between nodes. The association model of cross-platform advertising effect is established through the graph neural network, and the association of user's advertising interaction behavior on different platforms is analyzed through the graph neural network. Sp4: Real-time feedback and advertising strategy optimization. The main control module receives the associated data of each module and provides feedback and optimization. The system dynamically adjusts the display frequency of advertisements based on user behavior predictions. Based on the results of graph neural network analysis, the system can discover the synergy of cross-platform advertisements. Sp5: Continuous learning and model updating. The system updates data every day or every week, inputs new user behavior data into the time series model and graph neural network, recalculates advertising effect predictions and cross-platform correlations, and uses the latest data to dynamically adjust model parameters to ensure that the system can adapt to changing user behaviors and market dynamics.

8. The cross-platform advertising effect real-time digital tracking and analysis system according to claim 7, characterized in that: The data types collected in the data collection and preprocessing include user behavior data, platform data and timestamp data, and the data sources include client data, server data and user devices.

Images