Intelligent advertisement terminal based on big data and advertisement publishing method

Abstract

The invention discloses an intelligent advertisement terminal based on big data and an advertisement publishing method, and the intelligent advertisement terminal based on the big data comprises the following unit modules. A multi-source data acquisition unit; a data preprocessing and fusing unit; an intelligent analysis and decision unit; an advertisement display and interaction unit; a communication and feedback unit; the intelligent advertisement publishing method of the intelligent advertisement terminal based on big data comprises the following steps: S1, dynamically acquiring a strategy; s2, carrying out multi-modal data expansion; s3, adaptive feature extraction; s4, privacy protection enhancement; s5, performing hierarchical storage strategy and block chain storage expansion; s6, hybrid model construction and real-time decision support; s7, edge computing is coordinated and combined with intelligent cache optimization; by means of the terminal and the method, intelligent adaptation of the advertisement content, the user state and the environment can be achieved, the advertisement attraction and the user participation degree are remarkably enhanced, and the actual advertisement putting effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of advertising publishing, and in particular to a big data-based intelligent advertising terminal and an advertising publishing method. Background Art

[0002] With the rapid development of the Internet and digital technology, the advertising industry has ushered in tremendous changes, and the way of advertising delivery has gradually shifted towards intelligence and precision.

[0003] For example, the smart advertising terminal and advertising publishing method based on big data disclosed on the China Patent Network, whose patent publication number is "CN116137003A", mainly calculates multiple user visible areas of the smart advertising terminal based on terminal parameter information and terminal page area; predicts the gaze time of multiple user visible areas to obtain multiple predicted gaze times; divides the terminal page area into pages according to the multiple predicted gaze times to obtain target layout information; fills and renders the advertising information of the smart advertising terminal according to the target layout information and multiple target advertising information to obtain multiple target advertising terminal pages; obtains the advertising publishing order of multiple target advertising information, and scrolls and publishes and displays the multiple target advertising terminal pages according to the advertising publishing order. However, the advertising publishing terminal and method have disadvantages, namely, they cannot interact with many platforms and cannot receive more user data to build user models of the same type.

[0004] Based on this, the present invention proposes a new smart advertising terminal and advertising publishing method based on big data. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent advertising terminal and an advertising publishing method based on big data.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A smart advertising terminal based on big data, including the following unit modules:

[0008] Multi-source data collection unit: Equipped with edge computing equipment, it collects user behavior data, environmental data, and advertising-related data from multiple channels, while also monitoring the operating status of advertising terminal hardware in real time. It uses encryption technology to ensure data security during collection and transmission, and adopts blockchain-based distributed storage technology to store data in multiple nodes to prevent data loss. It also associates and tags various types of collected data to provide a comprehensive data foundation for subsequent processing.

[0009] Data preprocessing and fusion unit: performs feature extraction and dimensionality reduction on the collected data, and uses blockchain technology to record the entire data processing process to ensure that the data cannot be tampered with and is traceable;

[0010] Intelligent Analysis and Decision-Making Unit: This unit uses a logistic regression model to model user behavior, combined with semantic information provided by the knowledge graph to more accurately capture user behavior patterns. It also introduces a reinforcement learning mechanism, using advertising effectiveness as a reward signal to dynamically optimize advertising strategies and select appropriate advertising content and timing for different scenarios and user groups.

[0011] Advertisement display and interaction unit: This unit uses a recommendation algorithm based on user-item collaborative filtering to recommend advertisements to users. It combines environmental data and user behavior data acquired by the multi-source data acquisition unit to dynamically adjust the advertisement display style and interaction method. It also adaptively adjusts the advertisement display format based on factors such as the terminal device type, screen size, and network conditions, and transmits user interaction data to the communication and feedback unit in real time.

[0012] Communication and Feedback Unit: Responsible for data communication with external servers, uploading collected data and user interaction data, and receiving advertising resources and optimization strategies pushed by the server. An intelligent caching mechanism has been added to cache commonly used advertising resources and optimization strategies based on historical data and predicted demand from the ad display and interaction unit, ensuring continuity of ad delivery during poor network conditions. The unit also performs preliminary evaluation and screening of received optimization strategies.

[0013] As mentioned above, a smart advertising publishing method for a smart advertising terminal based on big data:

[0014] S1. Dynamic collection strategy: In the multi-source data collection phase, dynamic data collection rules are formulated based on the operating status of the advertising terminal hardware and environmental changes;

[0015] S2. Multimodal data expansion: In the multi-source data collection phase, technologies such as image recognition and speech recognition are introduced;

[0016] S3. Adaptive Feature Extraction: Conduct further in-depth analysis during the data preprocessing and fusion stages, build an adaptive feature extraction model based on deep learning, and automatically adjust feature extraction algorithms and parameters based on the characteristics and changing trends of different types of data to improve the accuracy and efficiency of feature extraction.

[0017] S4. Enhanced Privacy Protection: Building on federated learning and differential privacy technologies, homomorphic encryption technology is introduced to enable data computation and processing in encrypted form, further ensuring privacy during data preprocessing and fusion.

[0018] S5. Further adopt hierarchical storage strategy and blockchain storage expansion in the data preprocessing and fusion stage;

[0019] Tiered storage strategy: Data is divided into hot data, warm data, and cold data based on factors such as frequency of use and importance, and stored in storage media with different performance levels, reducing storage costs and improving data access efficiency.

[0020] Blockchain storage expansion: Optimizes blockchain-based distributed storage technology and adopts sharding chain technology to store data in different sub-chains, reducing the storage pressure of a single node and improving the scalability of data storage;

[0021] S6. Adopt hybrid model building and real-time decision support in the intelligent analysis and decision-making stage;

[0022] Hybrid model construction: By integrating logistic regression models, reinforcement learning mechanisms, and deep learning models, we build hybrid models to more accurately capture complex user behavior patterns and improve the accuracy and effectiveness of advertising strategies.

[0023] Real-time decision support: Build a real-time decision engine, combine real-time collected data with analysis models, and achieve instant response and adjustment of advertising strategies to meet dynamically changing advertising needs.

[0024] S6. During the ad display and interaction phase, we further utilize user historical behavior data, real-time status data, and knowledge graph information to build a personalized user interest model, achieve accurate personalized ad recommendations, and improve ad click-through rates and conversion rates.

[0025] S7: In the communication and feedback phase, edge computing and intelligent caching optimization strategies are used.

[0026] Edge computing collaboration: Strengthen the collaboration between edge computing devices and the communication stage, complete some data processing and analysis at the edge, reduce data upload volume, reduce network transmission pressure, and improve system response speed;

[0027] Intelligent cache optimization: Uses machine learning algorithms to predict the usage demand of advertising resources and optimization strategies, dynamically adjusts intelligent cache content, improves cache hit rate, and ensures smooth advertising delivery when the network is poor.

[0028] Preferably, the data preprocessing and fusion unit uses federated learning technology, and through federated learning combined with differential privacy technology, realizes data collaboration between different platforms or merchants under the premise of protecting data privacy, introduces knowledge graph technology, analyzes the correlation between data, constructs a knowledge graph in the advertising field, presents user behavior, advertising content, environmental factors and other information in a structured knowledge network, classifies and labels the processed data according to data type and application scenario, and provides accurate data support for subsequent units.

[0029] Preferably, the communication and feedback unit, in combination with the actual conditions such as local device performance and current advertising delivery effects, passes appropriate strategies to other unit modules for execution, regularly synchronizes data and interacts with other unit modules, and feeds back data transmission priorities and requirements to the multi-source data acquisition unit, promoting efficient collaboration among various unit modules to write further publishing and processing methods based on these materials.

[0030] Preferably, the intelligent analysis and decision-making unit uses federated learning to calculate the model gradient locally and perform gradient aggregation on the central server to achieve data collaboration between different platforms, and feed back the analysis results and decision basis to the data preprocessing and fusion unit to assist in optimizing data processing and fusion strategies.

[0031] Preferably, when the display screen temperature is too high or the processor load is continuously high, the frequency of collecting user behavior data and environmental data is automatically increased to ensure the integrity and real-time nature of key abnormal data.

[0032] Preferably, in the construction of the hybrid model and in the expansion of multimodal data, multimodal data of users in the advertising display scenario are collected, such as facial expressions and voice comments of users when watching advertisements, to enrich the data dimensions and provide more comprehensive information for user behavior analysis.

[0033] The present invention has the following beneficial effects:

[0034] 1. With the help of adaptive feature extraction technology and homomorphic encryption algorithm, secure ciphertext calculation is achieved in the data preprocessing stage, eliminating the risk of privacy leakage from the source; combined with blockchain distributed storage and hierarchical storage strategies, it eliminates the risk of data loss and reduces storage costs. At the same time, it greatly improves the automation and efficiency of data processing, ensuring the security and reliability of data throughout its life cycle.

[0035] 2. By integrating multiple intelligent algorithms to build a hybrid model, we can deeply explore user behavior patterns and accurately predict advertising strategies. At the same time, we can dynamically adjust advertising display and interaction methods based on multi-source data to achieve intelligent adaptation of advertising content to user status and environment, significantly enhance advertising appeal and user engagement, and improve the actual effect of advertising.

[0036] 3. Use federated learning and differential privacy technologies to break down data barriers and achieve efficient cross-platform data collaboration while protecting privacy; improve data transmission efficiency and network adaptability through edge computing collaboration and intelligent cache optimization, strengthen the dynamic feedback and collaboration mechanism between each unit module, comprehensively improve system operation efficiency, and build an efficient and collaborative smart advertising ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flowchart of the method for publishing the present invention;

[0038] Figure 2 This is a line graph of the annual data loss rate and data tampering rate in Example 1 of the present invention;

[0039] Figure 3 This is a line chart of the annual comprehensive model training accuracy and data utilization rate of Example 2 of the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0041] Reference Figure 1-3 , a smart advertising terminal based on big data, including the following unit modules:

[0042] Multi-source data collection unit: Equipped with edge computing equipment, it collects user behavior data, environmental data, and advertising-related data from multiple channels, while also monitoring the operating status of advertising terminal hardware in real time. It uses encryption technology to ensure data security during collection and transmission, and adopts blockchain-based distributed storage technology to store data in multiple nodes to prevent data loss. It also associates and tags various types of collected data to provide a comprehensive data foundation for subsequent processing.

[0043] Data preprocessing and fusion unit: performs feature extraction and dimensionality reduction on the collected data, and uses blockchain technology to record the entire data processing process to ensure that the data cannot be tampered with and is traceable;

[0044] Intelligent Analysis and Decision-Making Unit: This unit uses a logistic regression model to model user behavior, combined with semantic information provided by the knowledge graph to more accurately capture user behavior patterns. It also introduces a reinforcement learning mechanism, using advertising effectiveness as a reward signal to dynamically optimize advertising strategies and select appropriate advertising content and timing for different scenarios and user groups.

[0045] Advertisement display and interaction unit: This unit uses a recommendation algorithm based on user-item collaborative filtering to recommend advertisements to users. It combines environmental data and user behavior data acquired by the multi-source data acquisition unit to dynamically adjust the advertisement display style and interaction method. It also adaptively adjusts the advertisement display format based on factors such as the terminal device type, screen size, and network conditions, and transmits user interaction data to the communication and feedback unit in real time.

[0046] Communication and Feedback Unit: Responsible for data communication with external servers, uploading collected data and user interaction data, and receiving advertising resources and optimization strategies pushed by the server. An intelligent caching mechanism has been added to cache commonly used advertising resources and optimization strategies based on historical data and predicted demand from the ad display and interaction unit, ensuring continuity of ad delivery during poor network conditions. The unit also performs preliminary evaluation and screening of received optimization strategies.

[0047] As mentioned above, a smart advertising publishing method for a smart advertising terminal based on big data:

[0048] S1. Dynamic collection strategy: In the multi-source data collection phase, dynamic data collection rules are formulated based on the operating status of the advertising terminal hardware and environmental changes;

[0049] S2. Multimodal data expansion: In the multi-source data collection phase, technologies such as image recognition and speech recognition are introduced;

[0050] S3. Adaptive Feature Extraction: Conduct further in-depth analysis during the data preprocessing and fusion stages, build an adaptive feature extraction model based on deep learning, and automatically adjust feature extraction algorithms and parameters based on the characteristics and changing trends of different types of data to improve the accuracy and efficiency of feature extraction.

[0051] S4. Enhanced Privacy Protection: Building on federated learning and differential privacy technologies, homomorphic encryption technology is introduced to enable data computation and processing in encrypted form, further ensuring privacy during data preprocessing and fusion.

[0052] S5. Further adopt hierarchical storage strategy and blockchain storage expansion in the data preprocessing and fusion stage;

[0053] Tiered storage strategy: Data is divided into hot data, warm data, and cold data based on factors such as frequency of use and importance, and stored in storage media with different performance levels, reducing storage costs and improving data access efficiency.

[0054] Blockchain storage expansion: Optimizes blockchain-based distributed storage technology and adopts sharding chain technology to store data in different sub-chains, reducing the storage pressure of a single node and improving the scalability of data storage;

[0055] S6. Adopt hybrid model building and real-time decision support in the intelligent analysis and decision-making stage;

[0056] Hybrid model construction: By integrating logistic regression models, reinforcement learning mechanisms, and deep learning models, we build hybrid models to more accurately capture complex user behavior patterns and improve the accuracy and effectiveness of advertising strategies.

[0057] Real-time decision support: Build a real-time decision engine, combine real-time collected data with analysis models, and achieve instant response and adjustment of advertising strategies to meet dynamically changing advertising needs.

[0058] S6. During the ad display and interaction phase, we further utilize user historical behavior data, real-time status data, and knowledge graph information to build a personalized user interest model, achieve accurate personalized ad recommendations, and improve ad click-through rates and conversion rates.

[0059] S7: In the communication and feedback phase, edge computing and intelligent caching optimization strategies are used.

[0060] Edge computing collaboration: Strengthen the collaboration between edge computing devices and the communication stage, complete some data processing and analysis at the edge, reduce data upload volume, reduce network transmission pressure, and improve system response speed;

[0061] Intelligent cache optimization: Uses machine learning algorithms to predict the usage demand of advertising resources and optimization strategies, dynamically adjusts intelligent cache content, improves cache hit rate, and ensures smooth advertising delivery when the network is poor.

[0062] The data preprocessing and fusion unit uses federated learning technology, and through federated learning combined with differential privacy technology, realizes data collaboration between different platforms or merchants under the premise of protecting data privacy, introduces knowledge graph technology, analyzes the correlation between data, constructs a knowledge graph in the advertising field, and presents user behavior, advertising content, environmental factors and other information in a structured knowledge network. The processed data is classified and labeled according to data type and application scenario, providing accurate data support for subsequent units.

[0063] In the communication and feedback unit, based on the actual conditions such as local device performance and current advertising delivery effects, appropriate strategies are passed to other unit modules for execution, data synchronization and status interaction are regularly performed with other unit modules, and data transmission priorities and requirements are fed back to the multi-source data acquisition unit, promoting efficient collaboration among various unit modules to write further publishing and processing methods based on these materials.

[0064] In the intelligent analysis and decision-making unit, federated learning is used to calculate the model gradient locally and perform gradient aggregation on the central server, thereby achieving data collaboration between different platforms, feeding back the analysis results and decision basis to the data preprocessing and fusion unit to assist in optimizing the data processing and fusion strategy. When the display screen temperature is too high or the processor load is continuously high, the frequency of collecting user behavior data and environmental data is automatically increased to ensure the integrity and real-time nature of key abnormal data. In the construction of hybrid models and in the expansion of multimodal data, multimodal data of users in advertising display scenarios is collected, such as facial expressions and voice comments when watching advertisements, to enrich the data dimensions and provide more comprehensive information for user behavior analysis.

[0065] Example 1:

[0066] In the data preprocessing stage, feature extraction is performed on user behavior data and environmental data. Adaptive feature extraction is achieved using convolutional neural networks (CNN). Where m is the number of samples, n is the feature dimension, and after the convolution layer operation Y = f(W*X+b), where W is the convolution kernel weight matrix, * represents the convolution operation, b is the bias vector, and f is the activation function (such as the ReLU function f(x) = max(0,x)). Through multi-layer convolution and pooling operations, data features are automatically extracted.

[0067] In terms of privacy protection, homomorphic encryption technology is introduced and the CKKS homomorphic encryption algorithm is adopted. For the original data x, the public key pk is used to encrypt: c = Enc pk (x).

[0068] Perform data calculation in the ciphertext state, such as two encrypted data c1=Enc pk (x1) and c2 = Enc pk (x2) performs addition operation: c sum =c1+c2=Encc pk (x1+x2) multiplication operation: c mul =c1×c2=Enc pk (x1×x2) Finally, decrypt using the private key sk: x=Dec sk (c), thereby completing data preprocessing while ensuring data privacy.

[0069] The following data are obtained by applying Example 1 to the real advertising promotion method, as shown in Table 1:

[0070] Table 1: Comparison of annual data loss rate and data tampering rate

[0071] project April August December 16 months Data loss rate (%) 0.07％ 0.10％ 0.21％ 0.26％ Data tampering rate (%) 0.02％ 0.06％ 0.11％ 0.14％

[0072] Example 2:

[0073] In the intelligent analysis and decision-making unit, a hybrid model consisting of a logistic regression model, reinforcement learning mechanism and recurrent neural network (RNN) is constructed to optimize advertising delivery strategies.

[0074] For user behavior modeling, a logistic regression model is used to predict the probability of a user clicking an ad: P(y=1|x): Where x is the user behavior feature vector, θ is the model parameter vector, combined with RNN to the user behavior sequence, S=[s1,s2,…,s t ], modeling, the hidden layer state update formula is: h t =σ(W hh h t-1 +W xh x t +b h ), where h t is the hidden layer state at time t, W hh and W xh is the weight matrix b h is the bias vector σ for the activation function.

[0075] In the reinforcement learning mechanism, the advertising effect (click-through rate, conversion rate, etc.) is used as the reward signal r t Status t is the current user characteristics and advertising strategy, action a t The selected advertising strategy maximizes the cumulative discount reward. (γ is the discount factor, 0≤γ≤1)) to optimize the strategy and use the deep Q network (DQN) to learn the Q value function Q(s,a;θ):

[0076] Q(s t ,a t ;θ)←Q(s t ,a t ;θ)+α(r t +γmax a′ Q(s t+1 ,a′;θ - )-Q(s t ,a t ;θ)), where a is the learning rate θ - It is the target network parameter, and the dynamic optimization of the advertising delivery strategy is achieved through continuous iterative update of the strategy.

[0077] The following data are obtained by applying Example 2 to the real-world advertising promotion method, as shown in Table 2:

[0078] Table 2: Comparison of annual comprehensive model training accuracy and data utilization

[0079] project January April August December Model training accuracy% 98.1％ 93.6％ 92.8％ 91.8％ Data utilization % 86.8％ 87.6％ 84.2％ 89.1％

[0080] Embodiment 3;

[0081] In the advertising display and interaction unit, a recommendation algorithm based on user-item collaborative filtering is used, and advertising display is adjusted in combination with multi-source data.

[0082] In the traditional user-item collaborative filtering algorithm, the predicted score of user u for item i is calculated Adopt user-based collaborative filtering formula:

[0083]

[0084] in, and The average ratings of users u and v are N(u), and the set of users w similar to user u is uv is the similarity between users u and v, calculated using cosine similarity:

[0085]

[0086] Where I is the set of items that users u and v have jointly evaluated. Combining multi-source data, we introduce the environmental data weight ω e and user behavior data weight ω b Adjust the predicted score: Where e represents environmental data, b represents user behavior data, and f(e,b) is an adjustment function calculated from multi-source data. Based on the adjusted predicted scores, appropriate ads are recommended to users, and the ad display format and interaction methods are adaptively adjusted based on the terminal device.

[0087] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A smart advertising terminal based on big data, characterized in that: Includes the following unit modules: Multi-source data collection unit: Equipped with edge computing equipment, it collects user behavior data, environmental data, and advertising-related data from multiple channels, while also monitoring the operating status of advertising terminal hardware in real time. It uses encryption technology to ensure data security during collection and transmission, and adopts blockchain-based distributed storage technology to store data in multiple nodes to prevent data loss. It also associates and tags various types of collected data to provide a comprehensive data foundation for subsequent processing. Data preprocessing and fusion unit: performs feature extraction and dimensionality reduction on the collected data, and uses blockchain technology to record the entire data processing process to ensure that the data cannot be tampered with and is traceable; Intelligent Analysis and Decision-Making Unit: This unit uses a logistic regression model to model user behavior, combined with semantic information provided by the knowledge graph to more accurately capture user behavior patterns. It also introduces a reinforcement learning mechanism, using advertising effectiveness as a reward signal to dynamically optimize advertising strategies and select appropriate advertising content and timing for different scenarios and user groups. Advertisement display and interaction unit: This unit uses a recommendation algorithm based on user-item collaborative filtering to recommend advertisements to users. It combines environmental data and user behavior data acquired by the multi-source data acquisition unit to dynamically adjust the advertisement display style and interaction method. It also adaptively adjusts the advertisement display format based on factors such as the terminal device type, screen size, and network conditions, and transmits user interaction data to the communication and feedback unit in real time. Communication and Feedback Unit: Responsible for data communication with external servers, uploading collected data and user interaction data, and receiving advertising resources and optimization strategies pushed by the server. An intelligent caching mechanism has been added to cache commonly used advertising resources and optimization strategies based on historical data and predicted demand from the ad display and interaction unit, ensuring continuity of ad delivery during poor network conditions. The unit also performs preliminary evaluation and screening of received optimization strategies.

2. The big data-based smart advertising terminal according to claim 1, characterized in that: The data preprocessing and fusion unit uses federated learning technology, and through federated learning combined with differential privacy technology, realizes data collaboration between different platforms or merchants under the premise of protecting data privacy, introduces knowledge graph technology, analyzes the correlation between data, constructs a knowledge graph in the advertising field, and presents user behavior, advertising content, environmental factors and other information in a structured knowledge network. The processed data is classified and labeled according to data type and application scenario, providing accurate data support for subsequent units.

3. The big data-based smart advertising terminal according to claim 1, characterized in that: In the communication and feedback unit, based on the actual conditions such as local device performance and current advertising delivery effects, appropriate strategies are passed to other unit modules for execution, data synchronization and status interaction are regularly performed with other unit modules, and data transmission priorities and requirements are fed back to the multi-source data acquisition unit, promoting efficient collaboration among various unit modules to write further publishing and processing methods based on these materials.

4. The big data-based smart advertising terminal according to claim 1, characterized in that: The intelligent analysis and decision-making unit uses federated learning to calculate model gradients locally and aggregate gradients on a central server, enabling data collaboration between different platforms. The analysis results and decision basis are fed back to the data preprocessing and fusion unit to assist in optimizing data processing and fusion strategies.

5. A method for publishing smart advertisements by a smart advertising terminal based on big data according to any one of claims 1 to 4: S1. Dynamic collection strategy: In the multi-source data collection phase, dynamic data collection rules are formulated based on the operating status of the advertising terminal hardware and environmental changes; S2. Multimodal data expansion: In the multi-source data collection phase, technologies such as image recognition and speech recognition are introduced; S3. Adaptive Feature Extraction: Conduct further in-depth analysis during the data preprocessing and fusion stages, build an adaptive feature extraction model based on deep learning, and automatically adjust feature extraction algorithms and parameters based on the characteristics and changing trends of different types of data to improve the accuracy and efficiency of feature extraction. S4. Enhanced Privacy Protection: Building on federated learning and differential privacy technologies, homomorphic encryption technology is introduced to enable data computation and processing in encrypted form, further ensuring privacy during data preprocessing and fusion. S5. Further adopt hierarchical storage strategy and blockchain storage expansion in the data preprocessing and fusion stage; Tiered storage strategy: Data is divided into hot data, warm data, and cold data based on factors such as frequency of use and importance, and stored in storage media with different performance levels, reducing storage costs and improving data access efficiency. Blockchain storage expansion: Optimizes blockchain-based distributed storage technology and adopts sharding chain technology to store data in different sub-chains, reducing the storage pressure of a single node and improving the scalability of data storage; S6. Adopt hybrid model building and real-time decision support in the intelligent analysis and decision-making stage; Hybrid model construction: By integrating logistic regression models, reinforcement learning mechanisms, and deep learning models, we build hybrid models to more accurately capture complex user behavior patterns and improve the accuracy and effectiveness of advertising strategies. Real-time decision support: Build a real-time decision engine, combine real-time collected data with analysis models, and achieve instant response and adjustment of advertising strategies to meet dynamically changing advertising needs. S6. During the ad display and interaction phase, we further utilize user historical behavior data, real-time status data, and knowledge graph information to build a personalized user interest model, achieve accurate personalized ad recommendations, and improve ad click-through rates and conversion rates. S7: In the communication and feedback phase, edge computing and intelligent caching optimization strategies are used. Edge computing collaboration: Strengthen the collaboration between edge computing devices and the communication stage, complete some data processing and analysis at the edge, reduce data upload volume, reduce network transmission pressure, and improve system response speed; Intelligent cache optimization: Uses machine learning algorithms to predict the usage demand of advertising resources and optimization strategies, dynamically adjusts intelligent cache content, improves cache hit rate, and ensures smooth advertising delivery when the network is poor.

6. The intelligent advertisement publishing method according to claim 5, characterized in that: In the dynamic collection strategy, when the display screen temperature is too high or the processor load remains high, the collection frequency of user behavior data and environmental data will be automatically increased to ensure the integrity and real-time nature of key abnormal data.

7. The intelligent advertisement publishing method according to claim 5, characterized in that: In the construction of hybrid models and the expansion of multimodal data, multimodal data of users in advertising display scenarios are collected, such as facial expressions and voice comments when watching advertisements, to enrich data dimensions and provide more comprehensive information for user behavior analysis.

Images