Advertisement putting strategy optimization method and system based on preference data collaboration

Abstract

The invention discloses an advertisement putting strategy optimization method and system based on preference data collaboration, and relates to the technical field of advertisement putting, and the method comprises the steps: collecting the internal user data of a local advertisement putting platform, obtaining the external user data of a third-party platform through a safety data interface, carrying out the integration and preprocessing of the data, and obtaining an advertisement putting strategy; extracting explicit preference data and implicit preference data; a user-advertisement scoring matrix is constructed based on the preference data, and user preference vectors are generated after decomposition; acquiring copywriting text, image and category information of an advertisement to be put, constructing an advertisement feature expression model and generating an advertisement feature vector; calculating a matching degree score based on the user preference vector and the advertisement characteristic vector, and dynamically optimizing an advertisement putting strategy according to the score; according to the method, through comprehensive utilization of explicit and implicit preferences of the user, more accurate user interest modeling and advertisement matching are realized, and the advertisement putting effect and putting efficiency are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of advertising delivery technology, and in particular to a method and system for optimizing advertising delivery strategies based on preference data collaboration. Background Art

[0002] Preference data refers to data generated by collecting information about users' preferences or rankings of different options. It's typically presented in a comparative format rather than as absolute ratings, making it more representative of human subjective decision-making. This type of data comes from sources such as manual annotation, user behavior, and crowdsourcing platforms. Its core purpose is to help models learn "what humans prefer" rather than simply predicting the correct answer.

[0003] With the rapid development of the digital advertising market, advertisers and platforms face increasingly fierce competition. Achieving precise targeting within limited ad space resources and increasing click-through and conversion rates has become a key challenge. Traditional advertising strategies rely primarily on historical user behavior data for personalized recommendations, but these approaches often overlook differences in user preferences for different ad content, resulting in poor results. Furthermore, the limited user data available on a single platform makes it difficult to fully capture user interests, and the isolated nature of cross-platform data further limits the accuracy of ad recommendations. Summary of the Invention

[0004] In order to solve the above technical problems, a method and system for optimizing advertising delivery strategies based on preference data collaboration is provided. This technical solution solves the problem raised in the above background technology that with the rapid development of the digital advertising market, advertisers and platforms are facing increasingly fierce competition. How to achieve accurate delivery and improve user click-through rates and conversion rates with limited advertising resources has become a key challenge. Traditional advertising delivery strategies mainly rely on users' historical behavioral data for personalized recommendations, but such methods often ignore the differences in users' preferences for different advertising content, resulting in poor delivery effects. In addition, the user data of a single platform is limited, making it difficult to fully portray user interests, and the isolation of cross-platform data further limits the accuracy of advertising recommendations.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is: A method for optimizing advertising delivery strategies based on preference data collaboration, comprising: Obtain internal user data from local advertising platforms and external user data from third-party platforms through secure data interfaces; Integrate internal user data and external user data to obtain comprehensive user data, and pre-process the comprehensive user data to obtain explicit preference data and implicit preference data; Based on the explicit preference data and the implicit preference data, a user-advertisement rating matrix is ​​constructed, and the rating matrix is ​​subjected to matrix decomposition to obtain a user preference vector; Obtaining ad data from advertisers, extracting and analyzing the text, images, and structural category information from the ad data, and establishing an ad feature expression model. Based on the ad feature expression model, the ad feature vector is obtained. Calculate the matching degree between the user preference vector and the advertisement feature vector to obtain a matching score; Dynamically adjust advertising delivery strategies based on matching scores.

[0006] In an optional embodiment, the integrating of internal user data and external user data to obtain comprehensive user data, and preprocessing of the comprehensive user data to obtain explicit preference data and implicit preference data, specifically includes: Obtain internal user data of local advertising platforms; Collect external user data on third-party platforms by establishing a secure data interface connection based on an encrypted communication protocol with the third-party platform; Desensitize and standardize the external user data obtained from third-party platforms; Correlating and matching the processed external user data of the third-party platform with the internal user data of the advertising delivery platform, wherein the matching is based on user account, device identification or behavioral characteristics; Obtain comprehensive user data with a unified user ID; Preprocessing the comprehensive user data includes data cleaning, format unification, normalization, and behavior weight calculation; The pre-processed comprehensive user data is classified based on the active and passive nature of user behavior. The active behaviors of users such as rating, liking, collecting, and complaining about advertisements are identified as explicit preference data, and the passive behaviors such as click records, dwell time, conversion path, and page browsing depth during advertisement browsing are identified as implicit preference data.

[0007] In an optional embodiment, constructing a user-advertisement rating matrix based on explicit preference data and implicit preference data specifically includes: By normalizing the explicit preference data, the original scoring weights are set according to different behavior types; The original score weight values ​​are linearly normalized to map all explicit preference data to a unified [0,1] numerical range. The normalization formula is: ; Where, Represents a user About Advertising Explicit preference ratings of represents the user's original explicit behavior score, and are the minimum and maximum values ​​among all raw explicit behavior scores, respectively; By modeling behavioral intensity using implicit preference data; Based on implicit preference data, obtain ad click counts, page dwell time, browsing depth, and conversion event data; By integrating the data in a linear weighted manner, different weight coefficients are set for ad clicks, page dwell time, browsing depth, and conversion event data to obtain the user's implicit preference score for the ad. The calculation formula is: ; Where, represents the user's implicit preference rating for the advertisement, Represents a user About Advertising Click data rating, Indicates the length of stay data score, Indicates browsing depth data rating, Indicates conversion-related data scores. is the weight corresponding to each data item, satisfying ; The explicit preference rating and the implicit standard rating are weighted and integrated to construct the user-advertising rating matrix. , the expression formula of each element in the user-advertisement rating matrix R is: ; In the formula, the element Represents a user About Advertising Preference ratings, Represents a user About Advertising Explicit preference ratings of Represents a user About Advertising The implicit preference score of is the weighting coefficient, and its value range is [0,1].

[0008] In an optional embodiment, performing matrix decomposition processing on the rating matrix to obtain the user preference vector specifically includes: Decompose the rating matrix R into the user preference matrix and advertising feature matrix , satisfying the following approximate relationship: ; Where, is the user preference matrix with dimension m×k, is an advertising feature matrix with dimension n×k, where m is the number of users, n is the number of ads, and k is the dimension of the latent factor; The user preference matrix is ​​calculated by minimizing the loss function and advertising feature matrix To train: ; Where, is the user preference vector, is the advertisement latent vector, is the user preference matrix, is the advertising feature matrix, represents the rating matrix, A represents the sample set where the rating exists, and λ is the regularization coefficient.

[0009] In an optional embodiment, the method of obtaining the advertisement data to be delivered from the advertiser, extracting the copy text, images, and structural category information in the delivered advertisement data, analyzing the information, establishing an advertisement feature expression model, and obtaining an advertisement feature vector specifically includes: Based on the advertisement data to be placed, obtaining the text copy data to be placed, the image data to be placed, and the structural category information to be placed corresponding to the advertisement to be placed; Perform natural language processing on the text data corresponding to the advertisement to be placed, extract keywords and entity information related to the product function, purpose, selling point and emotional tendency, and convert them into text feature vectors through the word vector embedding model ; Perform image preprocessing, target detection, and deep feature extraction on the image data corresponding to the advertisement to be delivered. Use convolutional neural networks to extract multi-level image features and represent them as image feature vectors. ; Encode the category information of the structure to be delivered corresponding to the advertisement to be delivered and generate a category vector ; Fuse text feature vectors, image feature vectors, and category vectors to construct an advertisement feature expression model; According to the feature expression model of advertisement, the advertisement feature vector is obtained: ; Where, For advertising The characteristic vector of are the text, image and category feature vectors of the advertisement respectively; is the corresponding weighting coefficient, satisfying .

[0010] In an optional embodiment, the performing of matching calculation on the user preference vector and the advertisement characteristic vector to obtain a matching score specifically includes: User preference vector Vector with advertising characteristics Based on this, the matching score is obtained through the dot product scoring model; The expression formula of the dot product scoring model is: ; Where, For users and advertising The matching score, For users The preference vector of is the advertising feature vector The transpose of According to the matching score, when the matching score is higher than the set upper threshold, it is determined to be a high-potential user, and the system automatically increases the display frequency of the advertisement in the target user group, adjusts the exposure position to the home screen, top of the page, middle embedded position and the front position of the application homepage recommendation flow, improves real-time bidding, and prioritizes the allocation of high-exposure resources; when the matching score is in the middle range, it is determined to be a potential response customer, and the system will control the display time period and delivery duration of the advertisement, give priority to arranging it in the time window with higher conversion rate, and limit its daily delivery frequency and exposure quota; when the matching score is lower than the lower threshold, it is determined to be a low-relevance user, and the system will stop displaying the corresponding advertisement to it, reclaim allocated resources, and exclude the user from the subsequent redirection or expansion of the delivery population of the advertisement.

[0011] Furthermore, an advertisement delivery strategy optimization system based on preference data collaboration is proposed to implement the above-mentioned advertisement delivery strategy optimization method, which specifically includes: A data collection and fusion module, which is used to obtain internal user data of the local advertising platform and external user data of the third-party platform through a secure data interface; desensitize, standardize and correlate the internal and external user data to obtain comprehensive user data; and pre-process the comprehensive user data to obtain explicit preference data and implicit preference data; A feature modeling and vector generation module, which constructs a user-advertising rating matrix based on explicit and implicit preference data, performs matrix decomposition on the rating matrix, and generates a user preference vector; receives text, images, and category information of advertisements to be delivered, extracts features from the advertisement content, constructs an advertisement feature expression model, and generates an advertisement feature vector; The matching calculation and strategy optimization module calculates the matching degree according to the user preference vector and the advertisement feature vector, and dynamically adjusts the advertisement delivery strategy based on the matching degree score.

[0012] In an optional embodiment, the data acquisition and fusion module specifically includes: An internal data collection unit, configured to collect internal user data of the platform on which the advertisement is to be delivered; An external data acquisition unit, configured to obtain external user data of a third-party platform through a secure data interface; A data processing unit is provided, wherein the data processing unit fuses the internal user data and the external user data to obtain comprehensive user data, and pre-processes the comprehensive user data to obtain explicit preference data and implicit preference data.

[0013] In an optional embodiment, the feature modeling and vector generation module specifically includes: A rating matrix construction unit, which is used to normalize the explicit preference data and the implicit preference data and perform behavior intensity modeling to obtain explicit preference scores and implicit preference scores, and to construct a user advertising rating matrix based on the explicit preference scores and the implicit preference scores; A matrix decomposition unit is used to perform matrix decomposition on the rating matrix to obtain a user potential interest matrix and an advertising feature matrix, and obtain a user preference vector by minimizing the loss function training; The advertisement feature extraction and vector generation unit is used to analyze the text features, image features and category information in the advertisement data, construct an advertisement feature expression model, and obtain an advertisement feature vector.

[0014] In an optional embodiment, the matching calculation and strategy optimization module specifically includes: a matching degree calculation unit, wherein the matching degree calculation unit calculates a matching degree score based on the user preference vector and the advertisement feature vector using a dot product model method; A threshold determination unit, which is used to divide the matching score into three intervals: upper, medium, and lower, to determine the degree of association between the user and the advertisement; The delivery adjustment unit, the delivery strategy production unit is used to formulate a delivery strategy based on the matching interval results, including increasing the display frequency, adjusting the display position, controlling the frequency or stopping the display.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This solution proposes an advertising delivery strategy optimization method and system based on preference data collaboration. By integrating the internal user data of the local advertising delivery platform and the external user data of the third-party platform, a unified comprehensive user data is constructed. The data is pre-processed on the basis of desensitization and standardization conversion to identify explicit preference data and implicit preference data. A user-advertising rating matrix is ​​constructed based on the explicit and implicit preference data, and the user preference vector is obtained through matrix decomposition. This avoids the recommendation failure problem caused by the single data and insufficient understanding of preferences in traditional recommendation strategies, and fundamentally improves the relevance of advertising delivery and user acceptance.

[0016] This proposal proposes a method and system for optimizing advertising delivery strategies based on preference data collaboration. By combining ad text, images, and structural category information to construct an ad feature expression model, the system generates an ad feature vector. This vector then calculates the matching degree between the user preference vector and the ad feature vector. Based on this matching score, the system automatically adjusts display frequency, placement, and delivery time, effectively allocating advertising resources and avoiding ineffective exposure and budget waste. This method, without relying on manual rules, enables adaptive optimization of advertising strategies, effectively improving click-through rates, conversion rates, and overall delivery efficiency, providing advertising platforms with more real-time, intelligent, and revenue-oriented delivery strategy support. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of an advertising delivery strategy optimization method based on preference data collaboration proposed by the present invention; Figure 2 This is a flowchart for obtaining explicit preference data and implicit preference data in the present invention; Figure 3 This is a flowchart for obtaining the matching value score in the present invention; Figure 4 This is a system framework diagram of an advertising delivery strategy optimization system based on preference data collaboration proposed by the present invention. DETAILED DESCRIPTION

[0018] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0019] Reference Figure 1-Figure 4 As shown, a method for optimizing advertising delivery strategies based on preference data collaboration includes: Obtain internal user data from local advertising platforms and external user data from third-party platforms through secure data interfaces; Integrate internal user data and external user data to obtain comprehensive user data, and pre-process the comprehensive user data to obtain explicit preference data and implicit preference data; Based on explicit preference data and implicit preference data, a user-advertising rating matrix is ​​constructed, and the rating matrix is ​​subjected to matrix decomposition to obtain a user preference vector. Obtaining ad data from advertisers, extracting and analyzing the text, images, and structural category information from the ad data, and establishing an ad feature expression model. Based on the ad feature expression model, the ad feature vector is obtained. Calculate the matching degree between the user preference vector and the advertisement feature vector to obtain a matching score; Dynamically adjust advertising delivery strategies based on matching scores.

[0020] Furthermore, the internal user data and the external user data are integrated to obtain comprehensive user data, and the comprehensive user data is pre-processed to obtain explicit preference data and implicit preference data, specifically including: Obtain internal user data of local advertising platforms; Collect external user data on third-party platforms by establishing a secure data interface connection based on an encrypted communication protocol with the third-party platform; Desensitize and standardize the external user data obtained from third-party platforms; Specifically, after obtaining external user data from a third-party platform, the data is first desensitized and standardized. The core goal of desensitization is to ensure user privacy. In this step, the system encrypts data containing personal identity information, such as encrypting and storing sensitive information such as the user's real name and mobile phone number through an encryption algorithm. In addition, for data items that cannot be used directly, the system will adopt pseudo-anonymization processing, using random identifiers instead of real identity information, so as to ensure that the user's personal identity cannot be directly restored in the event of a data leak; Next, the system standardizes external user data to ensure consistency across different sources. During this process, the system uniformly converts formats such as dates, times, and numbers. For example, time fields are uniformly converted to the standard ISO 8601 format, and numerical data (such as purchase amounts and browsing time) are normalized according to preset intervals, ensuring that data from different platforms has the same dimensions and scale. Furthermore, text data is processed using a unified encoding method to ensure that it can be correctly parsed and applied in subsequent analysis.

[0021] Correlate and match the processed external user data of the third-party platform with the internal user data of the advertising platform, with the matching based on user account, device identifier or behavioral characteristics; It is understandable that once the desensitization and standardization conversion are completed, the system will begin to correlate and match the data. First, the system will match based on the user account. The system will determine whether the two data sources belong to the same user by comparing the user account information on the advertising platform and the third-party platform. For example, when the user uses the same account to log in on the advertising platform and the third-party platform, the system will merge these data to generate a unified portrait of the user. If the user account cannot be matched directly, the system will match based on the device identifier, which means that the system will compare the device ID or other device identification information from different platforms to try to identify whether the same device is used. For example, when a user uses the same device on multiple platforms, the system can use the device identifier to associate this data. If an accurate match is still not possible, the system will turn to a matching method based on user behavioral characteristics. Specifically, the system will analyze the user's behavioral characteristics on each platform and determine whether it is the same user by calculating the similarity of the behavior.

[0022] Obtain comprehensive user data with a unified user ID; By preprocessing comprehensive user data, including data cleaning, format unification, normalization and behavior weight calculation; The pre-processed comprehensive user data is classified based on the active and passive nature of user behavior. The active behaviors of users such as rating, liking, collecting, and complaining about advertisements are identified as explicit preference data, and the passive behaviors such as click records, dwell time, conversion path, and page browsing depth during advertisement browsing are identified as implicit preference data.

[0023] Furthermore, based on the explicit preference data and implicit preference data, a user-advertising rating matrix is ​​constructed, specifically including: By normalizing the explicit preference data, the original scoring weights are set according to different behavior types; The original score weight values ​​are linearly normalized to map all explicit preference data to a unified [0,1] numerical range. The normalization formula is: ; Where, Represents a user About Advertising Explicit preference ratings of represents the user's original explicit behavior score, and are the minimum and maximum values ​​among all raw explicit behavior scores, respectively; Specifically, for different types of user behaviors in explicit preference data, the differences in behavioral response intensity can be used as the basis for scoring weights. When analyzing different types of user interactions with advertisements, the system evaluates the subjectivity of their feedback and the depth of behavioral reach, and sets corresponding original scoring weight values. For example, behaviors with stronger subjective intentions (such as ad rating operations) are given higher original weights, while operations with lower operation thresholds and shallower behavioral reach (such as collecting or liking) are given relatively lower original weights. This scoring weight value serves as the initial quantitative basis for user preference tendencies, and is used for subsequent normalization and scoring matrix construction.

[0024] By modeling behavioral intensity using implicit preference data; Based on implicit preference data, obtain ad click counts, page dwell time, browsing depth, and conversion event data; By integrating the data in a linear weighted manner, different weight coefficients are set for ad clicks, page dwell time, browsing depth, and conversion event data to obtain the user's implicit preference score for the ad. The calculation formula is: ; Where, represents the user's implicit preference rating for the advertisement, Represents a user About Advertising Click data rating, Indicates the length of stay data score, Indicates browsing depth data rating, Indicates conversion-related data scores. is the weight corresponding to each data item, satisfying ; Specifically, when calculating the user's implicit preference score for an ad, different types of behavioral data are integrated through linear weighting. To improve the distinctiveness and practical orientation of the score, the system sets differentiated weight coefficients for different data based on the strength of each type of implicit behavior in representing the user's true interest. The number of ad clicks serves as a behavioral signal that indicates a user's initial interest. The behavioral threshold is low, and the weight coefficient is relatively small. The page dwell time reflects the user's attention to the ad content and can represent the content's attractiveness to a certain extent, so it is given a medium weight. The browsing depth reflects the user's extended exploration behavior under the ad content and represents the user's intention to actively obtain information. It has a slightly higher weight. The conversion event is the user's behavior of completing specific target actions such as registration, purchase, and jump. It has the strongest behavioral intention and commercial value, so it is set to the highest weight. For example, according to the rules of experience, the weight of click behavior is set to 0.1, the dwell time is 0.2, the browsing depth is 0.3, and the conversion behavior is 0.4. It can also be adaptively adjusted through data analysis and machine learning to optimize the delivery effect. The explicit preference rating and the implicit standard rating are weighted and integrated to construct the user-advertising rating matrix. , the expression formula of each element in the user-advertisement rating matrix R is: ; In the formula, the element Represents a user About Advertising Preference ratings, Represents a user About Advertising Explicit preference ratings of Represents a user About Advertising The implicit preference score of is the weighting coefficient, and its value range is [0,1].

[0025] Furthermore, the rating matrix is ​​decomposed to obtain the user preference vector, which specifically includes: Decompose the rating matrix R into the user preference matrix and advertising feature matrix , satisfying the following approximate relationship: ; Where, is the user preference matrix with dimension m×k, is an advertising feature matrix with dimension n×k, where m is the number of users, n is the number of ads, and k is the dimension of the latent factor; Specifically, when decomposing the user-advertising rating matrix R into a user preference matrix and an ad feature matrix, a latent factor dimension parameter k is introduced to limit the representation of users and ads in a low-dimensional feature space. This parameter k represents the dimensionality of the latent semantic factor, and its value determines the dimensionality of the user preference vector and the ad latent vector. By properly setting the k value, a balance is achieved between model complexity and prediction accuracy, ensuring that the matrix-decomposed representation of user preferences and ad features effectively extracts key features and exhibits good adaptability to new data.

[0026] The user preference matrix is ​​calculated by minimizing the loss function and advertising feature matrix To train: ; Where, is the user preference vector, is the advertisement latent vector, is the user preference matrix, is the advertising feature matrix, represents the rating matrix, A represents the sample set with ratings, and λ is the regularization coefficient to prevent overfitting.

[0027] Preferably, the user preference vector compresses the user's complex behavioral data and preference information into a low-dimensional vector representation, which can more comprehensively and accurately reflect the user's interest characteristics, significantly outperforming the traditional single behavior modeling method. Compared with relying only on a single dimension or a single type of behavioral data, the user preference vector integrates multiple explicit and implicit preferences, improves the relevance of personalization and recommendation, and at the same time improves computing efficiency through vectorized representation, facilitating large-scale real-time matching and dynamic optimization, thereby significantly enhancing the accuracy and intelligence of the advertising delivery system.

[0028] Furthermore, we obtain the advertisement data to be delivered from the advertiser, extract the copy text, images, and structural category information from the delivered advertisement data, analyze them, build an advertisement feature expression model, and obtain the advertisement feature vector, which specifically includes: Based on the advertisement data to be placed, obtaining the text copy data to be placed, the image data to be placed, and the structural category information to be placed corresponding to the advertisement to be placed; Perform natural language processing on the text data corresponding to the advertisement to be placed, extract keywords and entity information related to the product function, purpose, selling point and emotional tendency, and convert them into text feature vectors through the word vector embedding model ; Specifically, the system first performs Chinese word segmentation, part-of-speech tagging, and syntactic dependency analysis on the text copy data of the advertisement to be placed in order to identify the core content of the sentence; based on the constructed product knowledge graph and sentiment dictionary, the system identifies and extracts the keywords, product attribute words, usage scenario words, and positive and negative sentiment words in the copy, thereby extracting entities and semantic fragments that reflect the product functions, uses, selling points, and user emotional preferences; after completing the extraction of keywords and entity information, the system calls the pre-trained word vector embedding model to vectorize the above keywords and context information. By performing average pooling, weighted combination, or Transformer encoding on the word vector, a text feature vector of unified dimension is generated to represent the semantic information and communication focus of the advertisement copy, providing input support for the subsequent fusion calculation of the advertisement feature vector.

[0029] Perform image preprocessing, target detection, and deep feature extraction on the image data corresponding to the advertisement to be delivered. Use convolutional neural networks to extract multi-level image features and represent them as image feature vectors. ; It can be understood that the image data corresponding to the advertisement to be placed is first subjected to image preprocessing operations. Image preprocessing includes operations such as size unification, grayscale normalization and pixel normalization of the original image to improve the consistency and clarity of the image data; the preprocessed image is used by the target detection model to identify the product, person or brand area in the image, and the key area is intercepted for subsequent analysis; then, a convolutional neural network (CNN) model is constructed. The CNN model uses a network architecture composed of multiple convolution layers, pooling layers and nonlinear activation functions to extract the multi-level semantic features of the image layer by layer from the low-level texture edges, mid-level graphic contours and high-level semantic information of the image; finally, the output feature vector is used as the representation of the image for the subsequent acquisition of the advertising feature vector.

[0030] Encode the category information of the structure to be delivered corresponding to the advertisement to be delivered and generate a category vector ; Specifically, for the structural category information corresponding to the advertisements to be placed, the category data is first hierarchically parsed, and the category labels to which the advertisements belong are hierarchically split and standardizedly encoded; then, a one-hot encoding method is used for processing. First, a predefined category dictionary is established, and all possible advertisement category labels are uniformly numbered; then, according to the category position of each advertisement, a value of 1 is assigned to the corresponding dimension, and a value of 0 is assigned to the remaining dimensions, forming a sparse category vector. This category vector can accurately represent the specific classification to which the advertisement belongs, making it easier to obtain the advertisement feature vector in the subsequent process.

[0031] Fuse text feature vectors, image feature vectors, and category vectors to construct an advertisement feature expression model; According to the feature expression model of advertisement, the advertisement feature vector is obtained: ; Where, For advertising The characteristic vector of are the text, image and category feature vectors of the advertisement respectively; is the corresponding weighting coefficient, satisfying .

[0032] Furthermore, the user preference vector and the advertisement feature vector are matched to obtain a matching score, which specifically includes: User preference vector Vector with advertising characteristics Based on this, the matching score is obtained through the dot product scoring model; The expression formula of the dot product scoring model is: ; Where, For users and advertising The matching score, For users The preference vector of is the advertising feature vector The transpose of According to the matching score, when the matching score is higher than the set upper threshold, it is determined to be a high-potential user. The system automatically increases the display frequency of the advertisement in the target user group, adjusts the exposure position to the home screen, top of the page, middle embedded position and the front position of the application homepage recommendation flow, improves real-time bidding, and prioritizes the allocation of high-exposure resources; when the matching score is in the middle range, it is determined to be a potential response customer. The system will control the display time period and delivery duration of the advertisement, give priority to it in the time window with higher conversion rate, limit its daily delivery frequency and exposure quota to test the potential response of the user and balance resource input and return expectations; when the matching score is lower than the lower threshold, it is determined to be a low-relevance user. The system will stop showing the corresponding advertisement to it, recycle the allocated resources, and exclude the user from the subsequent redirection or expansion of the delivery population of the advertisement to prevent invalid exposure and budget waste.

[0033] Specifically, for example, the threshold is set to [0.5, 0.8]. When the matching score is greater than 0.8, the system identifies the user as a high-potential user, that is, the user has a high interest in the current advertising content or a high possibility of conversion. At this time, the system will perform advertising delivery enhancement operations; when the matching score is between [0.5, 0.8], the system regards the user as a potential response user, that is, the user has a certain interest in the advertisement but has not yet clearly expressed a strong preference. At this time, the system adopts a robust tentative delivery strategy; when the matching score is lower than 0.5, the system will determine the user as a low-relevance user, that is, the user lacks interest in the advertisement or conversion potential. At this time, the system will execute a negative screening strategy.

[0034] Furthermore, an advertisement delivery strategy optimization system based on preference data collaboration is proposed to implement the above-mentioned advertisement delivery strategy optimization method, which specifically includes: The data collection and fusion module is used to obtain internal user data from the local advertising platform and external user data from third-party platforms through a secure data interface; desensitize, standardize, and correlate internal and external user data to obtain comprehensive user data; and pre-process the comprehensive user data to obtain explicit and implicit preference data. The feature modeling and vector generation module constructs a user-advertising rating matrix based on explicit and implicit preference data, performs matrix decomposition on the rating matrix, and generates a user preference vector. It is used to receive the text, image, and category information of the advertisement to be delivered, extract features from the advertisement content, construct an advertisement feature expression model, and generate an advertisement feature vector. The matching calculation and strategy optimization module calculates the matching degree based on the user preference vector and the advertising feature vector, and dynamically adjusts the advertising delivery strategy based on the matching degree score.

[0035] Furthermore, the data collection and fusion module specifically includes: An internal data collection unit, which is used to collect internal user data of the platform where the advertisement is to be delivered; An external data acquisition unit, which is used to obtain external user data of a third-party platform through a secure data interface; The data processing unit integrates the internal user data and the external user data to obtain the comprehensive user data, and pre-processes the comprehensive user data to obtain the explicit preference data and the implicit preference data.

[0036] Furthermore, the feature modeling and vector generation module specifically includes: The scoring matrix construction unit is used to normalize explicit preference data and implicit preference data and perform behavioral intensity modeling to obtain explicit preference scores and implicit preference scores, and to construct a user advertising scoring matrix based on the explicit preference scores and implicit preference scores; Matrix decomposition unit: The matrix decomposition unit is used to decompose the rating matrix to obtain the user potential interest matrix and the advertising feature matrix, and obtain the user preference vector by minimizing the loss function training; The advertisement feature extraction and vector generation unit is used to analyze the text features, image features and category information in the advertisement data, build an advertisement feature expression model, and obtain an advertisement feature vector.

[0037] Furthermore, the matching calculation and strategy optimization module specifically includes: A matching degree calculation unit, which calculates a matching degree score based on the user preference vector and the advertisement feature vector using a dot product model method; A threshold judgment unit is used to divide the matching score into three categories: upper, medium, and lower, to determine the degree of association between the user and the advertisement; The delivery adjustment unit and delivery strategy production unit are used to formulate delivery strategies based on the matching interval results, including increasing display frequency, adjusting display positions, controlling frequency, or stopping display.

[0038] To sum up, the advantages of the present invention are: by integrating the internal user data of the local platform and the external user data of a third party, a unified comprehensive user data is constructed, explicit and implicit preferences are extracted and a user preference vector is generated, thereby avoiding the recommendation failure problem caused by the single data and insufficient understanding of preferences in traditional recommendation strategies, and at the same time integrating the advertising text, image and category information to construct an advertising feature vector, dynamically optimizing the display frequency, exposure position and delivery time based on the matching score, achieving intelligent advertising strategy adjustment with higher relevance and delivery efficiency, avoiding invalid exposure, and improving click-through rate and conversion effect.

[0039] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing advertising delivery strategies based on preference data collaboration, characterized in that: include: Obtain internal user data from local advertising platforms and external user data from third-party platforms through secure data interfaces; Integrate internal user data and external user data to obtain comprehensive user data, and pre-process the comprehensive user data to obtain explicit preference data and implicit preference data; Based on the explicit preference data and the implicit preference data, a user-advertisement rating matrix is ​​constructed, and the rating matrix is ​​subjected to matrix decomposition to obtain a user preference vector; Obtaining ad data from advertisers, extracting and analyzing the text, images, and structural category information from the ad data, and establishing an ad feature expression model. Based on the ad feature expression model, the ad feature vector is obtained. Calculate the matching degree between the user preference vector and the advertisement feature vector to obtain a matching score; Dynamically adjust advertising delivery strategies based on matching scores.

2. The method for optimizing advertising delivery strategies based on preference data collaboration according to claim 1, characterized in that: The integration of internal user data and external user data to obtain comprehensive user data, and preprocessing of the comprehensive user data to obtain explicit preference data and implicit preference data, specifically includes: Obtain internal user data of local advertising platforms; Collect external user data on third-party platforms by establishing a secure data interface connection based on an encrypted communication protocol with the third-party platform; Desensitize and standardize the external user data obtained from third-party platforms; Correlating and matching the processed external user data of the third-party platform with the internal user data of the advertising delivery platform, wherein the matching is based on user account, device identification or behavioral characteristics; Obtain comprehensive user data with a unified user ID; Preprocessing the comprehensive user data includes data cleaning, format unification, normalization, and behavior weight calculation; The pre-processed comprehensive user data is classified based on the active and passive nature of user behavior. The active behaviors of users such as rating, liking, collecting, and complaining about advertisements are identified as explicit preference data, and the passive behaviors such as click records, dwell time, conversion path, and page browsing depth during advertisement browsing are identified as implicit preference data.

3. The method for optimizing advertising delivery strategies based on preference data collaboration according to claim 1, characterized in that: The method of constructing a user-advertisement rating matrix based on explicit preference data and implicit preference data specifically includes: By normalizing the explicit preference data, the original scoring weights are set according to different behavior types; The original score weight values ​​are linearly normalized to map all explicit preference data to a unified [0,1] numerical range. The normalization formula is: ； Where, Represents a user About Advertising Explicit preference ratings of represents the user's original explicit behavior score, and are the minimum and maximum values ​​among all raw explicit behavior scores, respectively; By modeling behavioral intensity using implicit preference data; Based on implicit preference data, obtain ad click counts, page dwell time, browsing depth, and conversion event data; By integrating the data in a linear weighted manner, different weight coefficients are set for ad clicks, page dwell time, browsing depth, and conversion event data to obtain the user's implicit preference score for the ad. The calculation formula is: ； Where, represents the user's implicit preference rating for the advertisement, Represents a user About Advertising Click data rating, Indicates the length of stay data score, Indicates browsing depth data rating, Indicates conversion-related data scores. is the weight corresponding to each data item, satisfying ; The explicit preference rating and the implicit standard rating are weighted and integrated to construct the user-advertising rating matrix. , the expression formula of each element in the user-advertisement rating matrix R is: ； In the formula, the element Represents a user About Advertising Preference ratings, Represents a user About Advertising Explicit preference ratings of Represents a user About Advertising The implicit preference score of is the weighting coefficient, and its value range is [0,1].

4. The method for optimizing advertising delivery strategies based on preference data collaboration according to claim 1, characterized in that: The performing matrix decomposition processing on the rating matrix to obtain the user preference vector specifically includes: Decompose the rating matrix R into the user preference matrix and advertising feature matrix , satisfying the following approximate relationship: ； Where, is the user preference matrix with dimension m×k, is an advertising feature matrix with dimension n×k, where m is the number of users, n is the number of ads, and k is the dimension of the latent factor; The user preference matrix is ​​calculated by minimizing the loss function and advertising feature matrix To train: ； Where, is the user preference vector, is the advertisement latent vector, is the user preference matrix, is the advertising feature matrix, represents the rating matrix, A represents the sample set where the rating exists, and λ is the regularization coefficient.

5. The method for optimizing advertising delivery strategies based on preference data collaboration according to claim 1, characterized in that: The process of obtaining advertisement data to be placed from advertisers, extracting text, images, and structural category information from the advertisement data, analyzing the information, establishing an advertisement feature expression model, and obtaining an advertisement feature vector specifically includes: Based on the advertisement data to be placed, obtaining the text copy data to be placed, the image data to be placed, and the structural category information to be placed corresponding to the advertisement to be placed; Perform natural language processing on the text data corresponding to the advertisement to be placed, extract keywords and entity information related to the product function, purpose, selling point and emotional tendency, and convert them into text feature vectors through the word vector embedding model ; Perform image preprocessing, target detection, and deep feature extraction on the image data corresponding to the advertisement to be delivered. Use convolutional neural networks to extract multi-level image features and represent them as image feature vectors. ; Encode the category information of the structure to be delivered corresponding to the advertisement to be delivered and generate a category vector ; Fuse text feature vectors, image feature vectors, and category vectors to construct an advertisement feature expression model; According to the feature expression model of advertisement, the advertisement feature vector is obtained: ； Where, For advertising The characteristic vector of are the text, image and category feature vectors of the advertisement respectively; is the corresponding weighting coefficient, satisfying .

6. The method for optimizing advertising delivery strategies based on preference data collaboration according to claim 1, characterized in that: The matching calculation of the user preference vector and the advertisement feature vector to obtain a matching score specifically includes: User preference vector Vector with advertising characteristics Based on this, the matching score is obtained through the dot product scoring model; The expression formula of the dot product scoring model is: ； Where, For users and advertising The matching score, For users The preference vector of is the advertising feature vector The transpose of According to the matching score, when the matching score is higher than the set upper threshold, it is determined to be a high-potential user, and the system automatically increases the display frequency of the advertisement in the target user group, adjusts the exposure position to the home screen, top of the page, middle embedded position and the front position of the application homepage recommendation flow, improves real-time bidding, and prioritizes the allocation of high-exposure resources; when the matching score is in the middle range, it is determined to be a potential response customer, and the system will control the display time period and delivery duration of the advertisement, give priority to arranging it in the time window with higher conversion rate, and limit its daily delivery frequency and exposure quota; when the matching score is lower than the lower threshold, it is determined to be a low-relevance user, and the system will stop displaying the corresponding advertisement to it, reclaim allocated resources, and exclude the user from the subsequent redirection or expansion of the delivery population of the advertisement.

7. An advertising strategy optimization based on preference data collaboration The system, according to the method for optimizing advertising delivery strategies based on preference data collaboration according to claims 1-6, is characterized in that it specifically comprises: A data collection and fusion module, which is used to obtain internal user data of the local advertising platform and external user data of the third-party platform through a secure data interface; desensitize, standardize and correlate the internal and external user data to obtain comprehensive user data; and pre-process the comprehensive user data to obtain explicit preference data and implicit preference data; A feature modeling and vector generation module, which constructs a user-advertising rating matrix based on explicit and implicit preference data, performs matrix decomposition on the rating matrix, and generates a user preference vector; receives text, images, and category information of advertisements to be delivered, extracts features from the advertisement content, constructs an advertisement feature expression model, and generates an advertisement feature vector; The matching calculation and strategy optimization module calculates the matching degree according to the user preference vector and the advertisement feature vector, and dynamically adjusts the advertisement delivery strategy based on the matching degree score.

8. The system for optimizing advertising delivery strategies based on preference data collaboration according to claim 7, characterized in that: The data acquisition and fusion module specifically includes: An internal data collection unit, the internal data collection unit is used to collect internal user data of the platform to be advertised and obtain external user data of the third-party platform through a secure data interface; An external data acquisition unit, configured to obtain external user data of a third-party platform through a secure data interface; A data processing unit is provided, wherein the data processing unit fuses the internal user data and the external user data to obtain comprehensive user data, and pre-processes the comprehensive user data to obtain explicit preference data and implicit preference data.

9. The system for optimizing advertising delivery strategies based on preference data collaboration according to claim 7, characterized in that: The feature modeling and vector generation module specifically includes: A rating matrix construction unit, which is used to normalize the explicit preference data and the implicit preference data and perform behavior intensity modeling to obtain explicit preference scores and implicit preference scores, and to construct a user advertising rating matrix based on the explicit preference scores and the implicit preference scores; A matrix decomposition unit is used to perform matrix decomposition on the rating matrix to obtain a user potential interest matrix and an advertising feature matrix, and obtain a user preference vector by minimizing the loss function training; The advertisement feature extraction and vector generation unit is used to analyze the text features, image features and category information in the advertisement data, construct an advertisement feature expression model, and obtain an advertisement feature vector.

10. The system for optimizing advertising delivery strategies based on preference data collaboration according to claim 7, characterized in that: The matching calculation and strategy optimization module specifically includes: a matching degree calculation unit, wherein the matching degree calculation unit calculates a matching degree score based on the user preference vector and the advertisement feature vector using a dot product model method; A threshold determination unit, which is used to divide the matching score into three intervals: upper, medium, and lower, to determine the degree of association between the user and the advertisement; The delivery adjustment unit, the delivery strategy production unit is used to formulate a delivery strategy based on the matching interval results, including increasing the display frequency, adjusting the display position, controlling the frequency or stopping the display.

Images