Internet advertisement accurate putting method and system based on big data

Abstract

The invention discloses an internet advertisement accurate putting method and system based on big data, relates to the technical field of internet advertisements, and is used for solving the problem that the accuracy and dynamic adaptability of interest prediction are reduced due to the fact that advanced algorithms such as deep learning or reinforcement learning are not fully adopted by a model. User interaction features are captured in real time through a dynamic behavior analysis module, multi-dimensional data modeling is performed in combination with deep learning, and user interest prediction and real-time adjustment of an advertisement recommendation strategy are achieved. The dynamic behavior feature vector is fused with the time sequence and the spatial distribution feature, so that the prediction accuracy is improved; a multi-layer perceptron and a self-adaptive learning rate optimization algorithm are adopted to train a model, and generalization and response capabilities are enhanced; an advertisement adaptation degree score is calculated in combination with semantic similarity, and recommendation correlation is ensured; rapid feedback and strategy optimization are achieved through incremental learning, the defects of the prior art in the aspects of real-time performance and automation are overcome, and the advertisement putting efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of Internet advertising technology, and more specifically, to a method and system for precise Internet advertising delivery based on big data. Background Art

[0002] Precision internet advertising delivery technology leverages big data analytics and intelligent algorithms to personalize advertising. By integrating user behavior data, device information, and multi-dimensional features, it improves ad relevance and delivery efficiency. This technology helps optimize advertisers' return on investment (ROI) while improving user experience.

[0003] The existing technology has the following deficiencies: At present, static user portraits are constructed through user behavior data and basic information, and advertising push methods are implemented in combination with semantic understanding technology. However, the ability to respond to users' real-time dynamic behaviors is limited, and the comprehensive analysis of multi-dimensional data (such as time series, geographic location, etc.) is lacking, which may affect the timeliness of advertising recommendations. At the same time, although current technologies can integrate multi-channel data and generate delivery strategies, they have certain limitations in terms of the real-time nature of massive data processing and the complexity of algorithms. In addition, its user behavior model fails to fully adopt advanced algorithms such as deep learning or reinforcement learning, resulting in the need to improve the accuracy and dynamic adaptability of interest prediction. At the same time, cost control optimization relies on manual intervention, and the level of automation is low. Therefore, a method and system for precise Internet advertising delivery based on big data are proposed.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for precise Internet advertising delivery based on big data. By introducing a dynamic behavior analysis module to capture the interactive behavior characteristics of users in real time, and combining it with a deep learning algorithm to model and process multi-dimensional data, dynamic prediction of user interests and real-time adjustment of advertising recommendation strategies are completed, so as to solve the problems of insufficient real-time performance, limited multi-dimensional data integration capabilities and low degree of automation mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method and system for precise delivery of Internet advertisements based on big data, comprising the following steps: Step S1: Collecting real-time interactive behavior data of users on multiple platforms through a dynamic behavior analysis module, extracting user behavior pattern features and generating dynamic behavior feature vectors; Step S2: Use a deep learning model to fuse the dynamic behavior feature vector and the historical data analysis results to generate a user interest prediction model, and output the current user's interest preference distribution based on the user interest prediction model; Step S3: Match the interest preference distribution with the advertisement attributes in the advertisement resource pool, calculate the advertisement suitability score, and generate a preliminary advertisement recommendation list based on the order of the suitability score; Step S4: Collect the user's interactive behavior data on the preliminary advertisement recommendation list through the real-time feedback mechanism, update the dynamic behavior feature vector and readjust the user interest prediction model to optimize the advertisement recommendation strategy.

[0007] In a preferred embodiment, in step S1, the dynamic behavior analysis module obtains the user's real-time interactive behavior data from multiple data sources through distributed crawler technology. The real-time interactive behavior data includes clicks, browsing, dwell time, and cross-platform switching path information; the dynamic behavior feature vector is generated by weighted combination of the time series characteristics, spatial distribution characteristics, and behavior frequency characteristics of the user behavior.

[0008] In a preferred embodiment, in step S1, the historical data analysis results include the user's behavioral statistical characteristics over the past period of time, and the behavioral statistical characteristics include click-through rate, conversion rate, and interest category distribution within a specific time period; Access the historical database to obtain the user's historical behavior records, segment the historical behavior records using a sliding window algorithm, extract the behavior statistical features within each time period, and store the results as a historical behavior feature matrix; The historical behavior feature matrix is used to assist in the generation of dynamic behavior feature vectors.

[0009] In a preferred embodiment, in step S2, the deep learning model adopts a multi-layer perceptron structure, and the input layer receives the concatenation result of the dynamic behavior feature vector and the historical behavior feature matrix; The hidden layer extracts features and reduces the dimensionality of the input data through nonlinear activation functions, and the output layer generates the user interest preference distribution; The user interest preference distribution is represented by a probability distribution function, where the probability value of each interest category reflects the user's preference for that category.

[0010] In a preferred embodiment, in step S2, the training process of the user interest prediction model adopts an adaptive learning rate optimization algorithm to iteratively update the model parameters through a back-propagation mechanism; The training dataset consists of a historical behavior feature matrix and the corresponding user actual behavior labels. The validation dataset is used to evaluate the generalization ability of the model. After the model training is completed, the optimal parameter configuration is saved for subsequent reasoning.

[0011] In a preferred embodiment, in step S3, the advertisement attributes in the advertisement resource pool include advertisement category, target audience characteristics, advertisement display format, and advertisement budget limit; The ad suitability score is calculated using a weighted scoring formula, where the weight coefficients are dynamically adjusted based on the advertiser's priority needs. The preliminary ad recommendation list is generated by sorting the ad suitability scores from high to low.

[0012] In a preferred embodiment, in step S3, the calculation process of the advertisement suitability score combines the semantic similarity between the user interest preference distribution and the advertisement attributes, and the semantic similarity is calculated by the cosine similarity algorithm; The cosine similarity algorithm quantifies the matching degree between the user interest preference distribution vector and the advertisement attribute vector by calculating the cosine value of the angle between the user interest preference distribution vector and the advertisement attribute vector.

[0013] In a preferred embodiment, in step S4, the real-time feedback mechanism collects user interaction behavior data on the preliminary advertising recommendation list through tracking technology, and the interaction behavior data includes the number of clicks, stay time, and close operations; After preprocessing, the interactive behavior data generates a new dynamic behavior feature vector, which is used to update the user interest prediction model; The updated user interest prediction model recalculates the ad suitability score and generates an optimized ad recommendation list.

[0014] In a preferred embodiment, in step S4, the updating process of the user interest prediction model adopts an incremental learning algorithm to quickly process the newly added data through online learning; Incremental learning algorithms can make local adjustments to newly added data without retraining the entire model, thereby improving the model's real-time responsiveness.

[0015] A big data-based Internet advertising precision delivery system includes a dynamic behavior analysis module, a deep learning modeling module, an advertising matching module, and a real-time feedback optimization module, with signal connections between the modules; The dynamic behavior analysis module is used to collect real-time interactive behavior data of users on multiple platforms and generate dynamic behavior feature vectors; The deep learning modeling module is used to fuse dynamic behavior feature vectors and historical behavior feature matrices to generate a user interest prediction model and output the user interest preference distribution; The ad matching module is used to match the user's interest preference distribution with the ad attributes in the ad resource pool, calculate the ad suitability score and generate a preliminary ad recommendation list; The real-time feedback optimization module is used to collect user interaction behavior data on the preliminary advertising recommendation list, update the dynamic behavior feature vector and adjust the user interest prediction model, optimize the advertising recommendation strategy and generate an optimized advertising recommendation list.

[0016] Technical effects and advantages of the present invention: 1. The present invention uses a dynamic behavior analysis module to capture the interactive behavior characteristics of users in real time, combines deep learning algorithms to model and process multi-dimensional data, and realizes dynamic prediction of user interests and real-time adjustment of advertising recommendation strategies; the generation process of dynamic behavior feature vectors fully considers the time series characteristics and spatial distribution characteristics of user behavior, and improves the accuracy of user interest prediction; the deep learning model adopts a multi-layer perceptron structure and is trained through an adaptive learning rate optimization algorithm, which enhances the generalization ability and real-time response ability of the model; the calculation process of the advertising adaptability score combines the semantic similarity algorithm to ensure the relevance and accuracy of the advertising recommendation; the real-time feedback mechanism quickly processes user behavior data through an incremental learning algorithm, and further optimizes the dynamic adaptability of the advertising recommendation strategy. In summary, the present invention effectively solves the shortcomings of the existing technology in terms of real-time performance, multi-dimensional data integration and degree of automation, and meets the needs of advertisers and users for efficient and accurate advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a method for precise Internet advertising delivery based on big data according to the present invention.

[0018] Figure 2 This is a module structure framework diagram of a big data-based Internet advertising precision delivery system of the present invention.

[0019] Figure 3 This is a schematic diagram of the process of calculating the advertisement suitability score in a method for precise Internet advertisement delivery based on big data according to the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0021] Example 1 The present invention provides a method and system for accurate delivery of Internet advertisements based on big data. Figure 1 To the attached Figure 3 Specific embodiments of the present invention are described in detail.

[0022] First, based on Figure 2 As shown in the system structure diagram, the entire system consists of a dynamic behavior analysis module, a deep learning modeling module, an advertisement matching module, and a real-time feedback optimization module; These modules are interconnected and work together through data flows and logical relationships to complete the overall process from user behavior data collection to advertising recommendation strategy optimization; The dynamic behavior analysis module is responsible for collecting real-time interactive behavior data of users on multiple platforms and generating dynamic behavior feature vectors.

[0023] This module uses distributed crawler technology to obtain real-time user interaction behavior data from multiple data sources, including but not limited to e-commerce platforms, social media platforms, and search engines.

[0024] Specifically, the dynamic behavior analysis module first deploys data collection nodes on various platforms through tracking technology to capture information such as user clicks, browsing, dwell time, and cross-platform switching paths.

[0025] After preprocessing, these raw data are converted into a series of quantifiable behavioral indicators, such as time series characteristics, spatial distribution characteristics, and behavioral frequency characteristics.

[0026] In order to generate a dynamic behavior feature vector, the dynamic behavior analysis module adopts a weighted combination algorithm to fuse and calculate the above behavior indicators according to a certain weight coefficient, thereby obtaining a feature vector that can reflect the user's current behavior pattern.

[0027] The dynamic behavior analysis module also obtains the user's historical behavior records by accessing the historical database, and uses the sliding window algorithm to segment the historical behavior records, extract the behavioral statistical features within each time period, and finally generate a historical behavior feature matrix.

[0028] This matrix and the dynamic behavior feature vector together constitute the input data for the subsequent deep learning modeling module.

[0029] The deep learning modeling module receives the dynamic behavior feature vector and historical behavior feature matrix from the dynamic behavior analysis module, and performs fusion modeling on them to generate a user interest prediction model.

[0030] The core structure of the deep learning modeling module is the multi-layer perceptron (MLP). Its input layer receives the concatenation of the dynamic behavior feature vector and the historical behavior feature matrix. The hidden layer extracts features and reduces the dimensionality of the input data through a nonlinear activation function, and the output layer generates the user interest preference distribution.

[0031] The user interest preference distribution is represented by a probability distribution function, where the probability value of each interest category reflects the user's preference for that category.

[0032] In order to train the user interest prediction model, the deep learning modeling module adopts an adaptive learning rate optimization algorithm and iteratively updates the model parameters through the back-propagation mechanism.

[0033] The training dataset consists of a historical behavior feature matrix and the corresponding user actual behavior labels, and the validation dataset is used to evaluate the generalization ability of the model.

[0034] After model training is complete, the optimal parameter configuration is saved for subsequent inference. The output of the deep learning modeling module, namely the distribution of user interest preferences, is passed to the ad matching module as the basis for calculating the ad suitability score.

[0035] The ad matching module is responsible for matching user interest preference distribution with ad attributes in the ad resource pool and calculating ad suitability scores to generate a preliminary ad recommendation list.

[0036] The advertising resource pool stores a large amount of advertising data. Each piece of advertising data contains attribute information such as advertising category, target audience characteristics, advertising display format, and advertising budget limit.

[0037] The calculation process of the ad suitability score combines the semantic similarity between the user interest preference distribution and the ad attributes. The semantic similarity is calculated using the cosine similarity algorithm.

[0038] Specifically, the ad matching module first represents the user interest preference distribution as a vector and the ad attributes as a vector, and then quantifies the degree of matching between them by calculating the cosine value of the angle between the two vectors.

[0039] The weight coefficients in the calculation formula of the ad suitability score are dynamically adjusted according to the advertiser's demand priority. For example, for ads with higher budget limits, their weight coefficients will be appropriately increased.

[0040] The preliminary ad recommendation list is generated by sorting the ad suitability scores from high to low and passed to the real-time feedback optimization module.

[0041] The real-time feedback optimization module uses tracking technology to collect user interaction behavior data on the preliminary advertising recommendation list, including information such as the number of clicks, duration of stay, and closing operations.

[0042] These interactive behavior data are preprocessed to generate new dynamic behavior feature vectors, which are used to update the user interest prediction model.

[0043] The real-time feedback optimization module uses an incremental learning algorithm to quickly process new data and makes local adjustments to the user interest prediction model through online learning, thereby improving the model's real-time response capabilities.

[0044] The updated user interest prediction model recalculates the ad suitability score and generates an optimized ad recommendation list.

[0045] The optimized ad recommendation list is passed to the user end again for display, thus forming a complete closed-loop optimization process.

[0046] During the entire system operation, the dynamic behavior analysis module, deep learning modeling module, advertising matching module and real-time feedback optimization module work closely together through data flow and logical relationships.

[0047] The dynamic behavior feature vector and historical behavior feature matrix generated by the dynamic behavior analysis module provide basic data support for the deep learning modeling module; The user interest preference distribution generated by the deep learning modeling module provides key input for the ad matching module; The preliminary ad recommendation list generated by the ad matching module provides an optimization basis for the real-time feedback optimization module; The optimized advertising recommendation list generated by the real-time feedback optimization module further improves the accuracy and real-time performance of advertising recommendations.

[0048] Through this closed-loop design, the system can achieve dynamic prediction of user interests and real-time adjustment of advertising recommendation strategies.

[0049] Figure 1 A flowchart of a method for precise Internet advertising delivery based on big data provided by an embodiment of the present invention is shown; From step S1 to step S4, the overall steps from data collection to advertising recommendation strategy optimization are described in detail; Figure 2 It further clarifies the logical relationship and data flow direction between various modules in the system; Figure 3 The calculation process of the advertising suitability score is described in detail, and the matching calculation process between user interest preference distribution and advertising attributes is demonstrated.

[0050] By combining Figures 1 to 3 As well as the component numbers marked in the drawings, the specific implementation methods and operating principles of the present invention can be clearly understood.

[0051] In order to better enable relevant personnel in this technical field to fully understand and implement the present invention, the specific implementation principle of the present invention is further supplemented below with reference to a specific application scenario.

[0052] In actual application, it is assumed that an e-commerce platform hopes to use the Internet advertising precision delivery system based on big data provided by the present invention to optimize its advertising push effect.

[0053] The platform’s user behavior data sources include on-site browsing records, product click records, and off-site social media interaction data.

[0054] The following are the specific application scenarios and implementation steps of the operating principle: First, in the dynamic behavior analysis module, distributed crawler technology is used to obtain users' real-time interactive behavior data from multiple data sources.

[0055] For example, when a user browses a certain category of goods on an e-commerce platform, tracking technology will capture information such as the number of clicks, dwell time, and cross-platform switching paths of the user.

[0056] After preprocessing, these raw data are converted into a series of quantifiable behavioral indicators, such as time series characteristics (user activity in different time periods), spatial distribution characteristics (user access frequency on different pages or modules), and behavioral frequency characteristics (the number of times a specific operation occurs).

[0057] In order to generate dynamic behavior feature vectors, the dynamic behavior analysis module adopts a weighted combination algorithm to fuse and calculate the above behavior indicators according to certain weight coefficients.

[0058] For example, the weight coefficient of browsing behavior for high-value products will be appropriately increased to highlight their importance.

[0059] In addition, the dynamic behavior analysis module also obtains the user's historical behavior records by accessing the historical database, and uses the sliding window algorithm to segment the historical behavior records, extracts the behavioral statistical features within each time period, and finally generates a historical behavior feature matrix.

[0060] This process ensures that the dynamic behavior feature vector can fully reflect the user's current behavior pattern while taking into account the long-term trend of historical behavior.

[0061] Next, the deep learning modeling module receives the dynamic behavior feature vector and historical behavior feature matrix from the dynamic behavior analysis module and performs fusion modeling on them to generate a user interest prediction model.

[0062] In the specific implementation, the core structure of the deep learning modeling module is a multi-layer perceptron (MLP), whose input layer receives the concatenation result of the dynamic behavior feature vector and the historical behavior feature matrix.

[0063] The hidden layer extracts features and reduces the dimensionality of the input data through nonlinear activation functions, thereby capturing the potential patterns in user behavior.

[0064] The output layer generates the user interest preference distribution, where the probability value of each interest category reflects the user's preference for that category.

[0065] For example, if the user has frequently browsed electronic products recently, the probability value of the "electronic products" category will increase significantly.

[0066] In order to train the user interest prediction model, the deep learning modeling module adopts an adaptive learning rate optimization algorithm and iteratively updates the model parameters through the back-propagation mechanism.

[0067] During the training process, the model will continuously adjust the parameter configuration based on the historical behavior feature matrix and the corresponding user actual behavior labels to improve prediction accuracy.

[0068] After training is completed, the optimal parameter configuration is saved for subsequent inference.

[0069] Subsequently, the ad matching module matches the user interest preference distribution with the ad attributes in the ad resource pool and calculates the ad suitability score to generate a preliminary ad recommendation list.

[0070] In a specific implementation, a large amount of advertising data is stored in the advertising resource pool. Each piece of advertising data contains attribute information such as advertising category, target audience characteristics, advertising display format, and advertising budget limit.

[0071] The calculation process of the ad suitability score combines the semantic similarity between the user interest preference distribution and the ad attributes. The semantic similarity is calculated using the cosine similarity algorithm.

[0072] For example, if the user interest preference distribution vector is [0.1, 0.7, 0.2], indicating that the user has a high preference for "electronic products", and an advertisement attribute vector is [0.2, 0.6, 0.2], then the cosine similarity value between the two is high, indicating that the advertisement is highly matched with the user's interests.

[0073] The weight coefficients in the calculation formula of the advertisement suitability score are dynamically adjusted according to the advertiser's demand priority.

[0074] For example, for advertisements with higher budget limits, their weight coefficients will be appropriately increased to ensure the advertiser's input-output ratio.

[0075] The preliminary ad recommendation list is generated by sorting the ad suitability scores from high to low and passed to the real-time feedback optimization module.

[0076] Finally, the real-time feedback optimization module uses tracking technology to collect user interaction behavior data on the preliminary advertising recommendation list, including information such as the number of clicks, duration of stay, and closing operations.

[0077] These interactive behavior data are preprocessed to generate new dynamic behavior feature vectors, which are used to update the user interest prediction model.

[0078] The real-time feedback optimization module uses an incremental learning algorithm to quickly process new data and makes local adjustments to the user interest prediction model through online learning.

[0079] For example, if a user shows a high click-through rate for an advertisement, the model will accordingly increase the weight of the category to which the advertisement belongs, thereby improving the model's real-time responsiveness.

[0080] The updated user interest prediction model recalculates the ad suitability score and generates an optimized ad recommendation list.

[0081] The optimized ad recommendation list is passed to the user end again for display, thus forming a complete closed-loop optimization process.

[0082] During the entire system operation, the dynamic behavior analysis module, deep learning modeling module, advertising matching module and real-time feedback optimization module work closely together through data flow and logical relationships.

[0083] The dynamic behavior feature vector and historical behavior feature matrix generated by the dynamic behavior analysis module provide basic data support for the deep learning modeling module. The user interest preference distribution generated by the deep learning modeling module provides key input for the advertising matching module. The preliminary advertising recommendation list generated by the advertising matching module provides an optimization basis for the real-time feedback optimization module, and the optimized advertising recommendation list generated by the real-time feedback optimization module further improves the accuracy and real-time performance of advertising recommendations.

[0084] Through this closed-loop design, the system can achieve dynamic prediction of user interests and real-time adjustment of advertising recommendation strategies.

[0085] By combining the steps and principles of the above specific application scenarios, we can clearly understand how the present invention can achieve efficient and accurate advertising delivery in practice.

[0086] The dynamic behavior analysis module captures user behavior data in real time through distributed crawler technology and embedding technology. The deep learning modeling module generates a user interest prediction model through a multi-layer perceptron structure and an adaptive learning rate optimization algorithm. The advertising matching module calculates the advertising suitability score through a cosine similarity algorithm. The real-time feedback optimization module quickly processes user behavior data through an incremental learning algorithm.

[0087] These technical means work together to ensure the relevance, accuracy and real-time nature of advertising recommendations, effectively addressing the shortcomings of existing technologies in terms of real-time performance, multi-dimensional data integration and automation.

[0088] Example 2 See also Figure 1 and Figure 3 , a method for accurately delivering Internet advertisements based on big data, comprising the following steps: Step S1: Collecting real-time interactive behavior data of users on multiple platforms through a dynamic behavior analysis module, extracting user behavior pattern features and generating dynamic behavior feature vectors; Step S2: Use a deep learning model to fuse the dynamic behavior feature vector and the historical data analysis results to generate a user interest prediction model, and output the current user's interest preference distribution based on the user interest prediction model; Step S3: Match the interest preference distribution with the advertisement attributes in the advertisement resource pool, calculate the advertisement suitability score, and generate a preliminary advertisement recommendation list based on the order of the suitability score; Step S4: Collecting user interaction behavior data on the preliminary advertising recommendation list through a real-time feedback mechanism, updating the dynamic behavior feature vector and re-adjusting the user interest prediction model to optimize the advertising recommendation strategy; The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0089] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0090] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0091] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0092] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0093] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0094] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0096] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0097] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0098] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0099] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for precise internet advertising delivery based on big data, characterized by: The following steps are involved: Step S1: Collecting real-time interactive behavior data of users on multiple platforms through a dynamic behavior analysis module, extracting user behavior pattern features and generating dynamic behavior feature vectors; Step S2: Use a deep learning model to fuse the dynamic behavior feature vector and the historical data analysis results to generate a user interest prediction model, and output the current user's interest preference distribution based on the user interest prediction model; Step S3: Match the interest preference distribution with the advertisement attributes in the advertisement resource pool, calculate the advertisement suitability score, and generate a preliminary advertisement recommendation list based on the order of the suitability score; Step S4: Collect the user's interactive behavior data on the preliminary advertisement recommendation list through the real-time feedback mechanism, update the dynamic behavior feature vector and readjust the user interest prediction model to optimize the advertisement recommendation strategy.

2. The method for precise internet advertising delivery based on big data according to claim 1, characterized in that: In step S1, the dynamic behavior analysis module obtains the user's real-time interactive behavior data from multiple data sources through distributed crawler technology. The real-time interactive behavior data includes clicks, browsing, dwell time, and cross-platform switching path information; the dynamic behavior feature vector is generated by weighted combination of the time series characteristics, spatial distribution characteristics, and behavior frequency characteristics of the user behavior.

3. The method for precise internet advertising delivery based on big data according to claim 2, characterized in that: In step S1, the historical data analysis results include the user's behavioral statistical characteristics over the past period of time, and the behavioral statistical characteristics include click-through rate, conversion rate, and interest category distribution within a specific time period; Access the historical database to obtain the user's historical behavior records, segment the historical behavior records using a sliding window algorithm, extract the behavior statistical features within each time period, and store the results as a historical behavior feature matrix; The historical behavior feature matrix is used to assist in the generation of dynamic behavior feature vectors.

4. The method for precise internet advertising delivery based on big data according to claim 3, characterized in that: In step S2, the deep learning model adopts a multi-layer perceptron structure, and the input layer receives the concatenation result of the dynamic behavior feature vector and the historical behavior feature matrix; The hidden layer extracts features and reduces the dimensionality of the input data through nonlinear activation functions, and the output layer generates the user interest preference distribution; The user interest preference distribution is represented by a probability distribution function, where the probability value of each interest category reflects the user's preference for that category.

5. The method for precise internet advertising delivery based on big data according to claim 4, characterized in that: In step S2, the training process of the user interest prediction model adopts an adaptive learning rate optimization algorithm to iteratively update the model parameters through the back-propagation mechanism; The training dataset consists of a historical behavior feature matrix and the corresponding user actual behavior labels. The validation dataset is used to evaluate the generalization ability of the model. After the model training is completed, the optimal parameter configuration is saved for subsequent reasoning.

6. The method for precise internet advertising delivery based on big data according to claim 1, characterized in that: In step S3, the advertisement attributes in the advertisement resource pool include advertisement category, target audience characteristics, advertisement display format, and advertisement budget limit; The ad suitability score is calculated using a weighted scoring formula, where the weight coefficients are dynamically adjusted based on the advertiser's priority needs. The preliminary ad recommendation list is generated by sorting the ad suitability scores from high to low.

7. The method for precise internet advertising delivery based on big data according to claim 6, characterized in that: In step S3, the calculation process of the advertisement suitability score combines the semantic similarity between the user interest preference distribution and the advertisement attributes, and the semantic similarity is calculated using the cosine similarity algorithm; The cosine similarity algorithm quantifies the matching degree between the user interest preference distribution vector and the advertisement attribute vector by calculating the cosine value of the angle between the user interest preference distribution vector and the advertisement attribute vector.

8. The method for precise internet advertising delivery based on big data according to claim 1, characterized in that: In step S4, the real-time feedback mechanism collects user interaction behavior data on the preliminary ad recommendation list through tracking technology. The interaction behavior data includes the number of clicks, stay time, and closing operations; After preprocessing, the interactive behavior data generates a new dynamic behavior feature vector, which is used to update the user interest prediction model; The updated user interest prediction model recalculates the ad suitability score and generates an optimized ad recommendation list.

9. The method for precise internet advertising delivery based on big data according to claim 8, characterized in that: In step S4, the updating process of the user interest prediction model adopts an incremental learning algorithm to quickly process the newly added data through online learning; Incremental learning algorithms can make local adjustments to newly added data without retraining the entire model, thereby improving the model's real-time responsiveness.

10. A system for accurately delivering internet advertisements based on big data, for implementing the method for accurately delivering internet advertisements based on big data according to any one of claims 1 to 9, characterized in that: It includes a dynamic behavior analysis module, a deep learning modeling module, an ad matching module, and a real-time feedback optimization module, with signal connections between each module; The dynamic behavior analysis module is used to collect real-time interactive behavior data of users on multiple platforms and generate dynamic behavior feature vectors; The deep learning modeling module is used to fuse dynamic behavior feature vectors and historical behavior feature matrices to generate a user interest prediction model and output the user interest preference distribution; The ad matching module is used to match the user's interest preference distribution with the ad attributes in the ad resource pool, calculate the ad suitability score and generate a preliminary ad recommendation list; The real-time feedback optimization module is used to collect user interaction behavior data on the preliminary advertising recommendation list, update the dynamic behavior feature vector and adjust the user interest prediction model, optimize the advertising recommendation strategy and generate an optimized advertising recommendation list.

Images