Big data-based directional and accurate delivery system and method for media promotion

Abstract

The invention relates to the technical field of directional delivery, in particular to a media promotion directional precise delivery system and method based on big data, and the system comprises a user interest mining module, a content correlation analysis module, a heterogeneous network optimization module, a user behavior insight module, an advertisement matching strategy module and a precise delivery effect monitoring module. According to the method, the advertisement content matching accuracy is further improved by analyzing user browsing contents and records through a big data technology, effectively extracting keywords, identifying interest points and predicting the behavior trend of the user, it is ensured that advertisements can be accurately put to directional target users every time, and the advertisement effect is monitored in real time, so that the advertisement quality is improved. The advertising strategy is adjusted according to the user feedback, so that the advertising efficiency and the user satisfaction degree are remarkably improved, accurate matching and advertising of the advertising content are achieved through deep analysis of user behaviors and interests, the pertinence and effectiveness of advertisements are greatly enhanced, meanwhile, the user experience is optimized, and interference of irrelevant advertisements is avoided.

Description

Technical Field

[0001] The present invention relates to the field of targeted delivery technology, and in particular to a system and method for targeted and precise delivery of media promotion based on big data. Background Art

[0002] The field of targeted delivery technology is a field that involves using data analysis to accurately identify target audiences and display relevant advertising content. By analyzing massive amounts of data (big data) to identify key factors such as potential customers' interests, purchasing habits, demographic information, etc., more efficient advertising can be achieved. Not only does it increase the ROI (return on investment) of advertising activities, but it also improves user experience because the advertising content that users are exposed to is closer to their actual needs and interests.

[0003] Among them, the media promotion targeted precision delivery system based on big data is a highly technical platform. The core is to use big data technology to analyze user behavior and preferences, so as to achieve accurate matching and delivery of advertising content. The main purpose is to improve advertising effects, reduce invalid exposure, and maximize the efficiency of advertising budget by accurately positioning the target audience. Through in-depth analysis of user data, the system can predict user behavior and push the most relevant advertising content, in order to achieve the dual effects of improving advertising conversion rate and enhancing user satisfaction.

[0004] Traditional targeted delivery technology lacks sufficiently in-depth analysis of user behavior and interests, resulting in low accuracy in ad matching and failure to fully utilize the potential of big data to optimize ad delivery strategies. Superficial data analysis leads to a disconnect between ad content and actual user needs and interests, increasing ineffective exposure and reducing ROI. Traditional systems push ads to non-target groups, resulting in a decline in user experience, and users ignoring or blocking ads, resulting in a waste of budget and a negative impact on brand image. The lack of an effective feedback adjustment mechanism makes it difficult to optimize ad strategies in a timely manner based on market feedback, further reducing the effectiveness and efficiency of ad delivery. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a media promotion targeted and precise delivery system and method based on big data.

[0006] To achieve the above purpose, the present invention adopts the following technical scheme: a media promotion directional and precise delivery system based on big data includes: The user interest mining module is based on user online behavior data and uses big data technology to analyze user browsing content and records, extract keywords, identify points of interest, classify positive and negative emotions, and generate interest preference feature sets; The content association analysis module analyzes the interest point relationship based on the interest preference feature set, classifies similar interest groups, identifies interest trends, and generates an interest association map; The heterogeneous network optimization module uses a graph algorithm to adjust the connection mode of nodes and edges in the heterogeneous network based on the interest association graph, uses a meta-path selection method to determine the path between nodes of different types, and generates an optimized user interest network; The user behavior insight module models the user behavior sequence based on the optimized user interest network, analyzes the potential trends and motivations of the behavior, and generates a behavior trend prediction model; The advertisement matching strategy module screens matching advertisement contents, sorts user group suitability, and generates a targeted advertisement content list based on the behavior trend prediction model; The precise delivery effect monitoring module monitors the advertising delivery effect based on the targeted advertising content list, adjusts the strategy according to user interaction, and generates the advertising precise delivery effect analysis result.

[0007] As a further solution of the present invention, the interest preference feature set includes user topic preferences, active time periods and interaction frequencies, the interest association map includes the correlation between topics, interest evolution paths and group preference trends, the optimized user interest network includes node correlation areas, key propagation paths and information aggregation efficiency highlights, the behavior trend prediction model includes behavior occurrence probability, trend turning points and potential interest triggering factors, the targeted advertising content list includes advertising topic suitability, target user matching and content innovation points, and the advertising precision delivery effect analysis results include click-through rate highlights, user satisfaction changes and delivery cost-effectiveness ratio.

[0008] As a further solution of the present invention, the user interest mining module includes: The content analysis submodule uses big data technology to analyze users' browsing history based on their online behavior data, extract keywords with increasing frequency, identify users' points of interest through keywords, and generate points of interest identification results; The emotion classification submodule analyzes the user's comments and feedback on the interest points based on the interest point recognition results, distinguishes the emotional tendencies, and classifies the interest points and emotional tendencies to obtain the emotion classification results; The feature set integration submodule associates and integrates the user's interest points with the emotional tendency based on the emotion classification result, and generates an interest preference feature set by analyzing the user's interest and emotional feedback.

[0009] As a further solution of the present invention, the content association analysis module includes: The interest point analysis submodule analyzes the user behavior data based on the interest preference feature set, identifies the user's interest points, extracts the key features of the interest points, and generates interest point keywords; The interest group classification submodule divides users into groups based on the interest point keywords, divides users into differentiated groups according to the similarity of interest points, and generates similar interest group classifications; The trend identification submodule is based on the classification of the same interest groups and uses big data analysis technology to identify the development and change trends of the interest points of multiple groups and generate an interest association map.

[0010] As a further solution of the present invention, the heterogeneous network optimization module includes: The network structure adjustment submodule is based on the interest association graph. If the connection between nodes does not meet the optimization target, the connection mode of nodes and edges is reconfigured to generate an adjusted network structure. The path selection submodule is based on the adjusted network structure, and if there are multiple paths connecting nodes of different types, a meta-path selection method is used to determine the optimized path and generate an optimized path set; The network efficiency optimization submodule is based on the optimized path set. If the information transmission efficiency in the network is delayed, the network efficiency optimization submodule generates an optimized user interest network by adjusting the weight distribution of the paths and nodes.

[0011] As a further solution of the present invention, the user behavior insight module includes: The behavior sequence modeling submodule collects user behavior sequence data based on the optimized user interest network. If the user activity presents a periodic pattern, it is marked as a periodic behavior, and the user's behavior sequence is modeled to generate a user behavior model. The trend analysis submodule applies a convolutional neural network based on the user behavior model to analyze the time series of the behavior data. If the behavior trend increases, it is marked as a growth trend, the potential trend of the behavior is identified, and a potential trend analysis result is generated; The convolutional neural network algorithm follows the formula:

[0012]

[0013]

[0014] in, is the output of the convolutional layer, is the original weight of the convolution kernel, is the activation function, is the dynamic adjustment of weight, is the weight adjustment coefficient, The reference time window The input data, is the original bias term, is the dynamic adjustment of the bias term, is the bias adjustment coefficient, is the original weight of the fully connected layer, is the dynamic adjustment amount of the fully connected layer weight, is the weight adjustment coefficient of the fully connected layer, is the original bias term of the fully connected layer, is the dynamic adjustment of the bias term of the fully connected layer, is the bias adjustment coefficient of the fully connected layer, is the output of the fully connected layer, is the activation function used for the fully connected layer, is the result of trend analysis. is an improved trend identification function, and New parameters introduced for trend analysis; The motivation identification submodule analyzes the motivation behind the user behavior based on the potential trend analysis results. If the target behavior is related to the user's interest point, it is identified as interest-driven behavior. The user's behavioral motivation is identified from trends and behavioral patterns, and a behavioral trend prediction model is generated.

[0015] As a further solution of the present invention, the advertisement matching strategy module includes: The advertisement content screening submodule analyzes the user behavior data based on the behavior trend prediction model, identifies the advertisement content associated with the user's interests, selects advertisements that match the user's behavior pattern, and generates a matching advertisement set; The suitability ranking submodule evaluates the suitability of the advertisement content based on the matching advertisement set, ranks the advertisement content according to the suitability with reference to the diversity of user interests and the frequency of behaviors, and generates a ranked advertisement list; The content list generation submodule uses big data analysis technology to screen advertising content associated with user interests and behavior trends based on the sorted advertising list and generates a targeted advertising content list.

[0016] As a further solution of the present invention, the precise delivery effect monitoring module includes: The advertising performance monitoring submodule collects the number of advertisement displays, click-through rates, and initialization responses of users based on the targeted advertisement content list, establishes performance indicators for the advertisements, and generates advertising performance monitoring data; The conversion rate analysis submodule analyzes the user's purchase behavior or form filling after clicking on the advertisement based on the advertisement performance monitoring data, identifies the conversion rate of multiple advertisement contents, and generates a conversion rate analysis result; The delivery strategy optimization submodule evaluates the effects of multiple advertising contents based on the conversion rate analysis results, proposes adjustment plans for inefficient advertisements, optimizes the target audience settings for precise advertising delivery, and obtains analysis results of the precise advertising delivery effects.

[0017] A method for targeted and precise delivery of media promotion based on big data, wherein the method for targeted and precise delivery of media promotion based on big data is executed based on the above-mentioned targeted and precise delivery system for media promotion based on big data, and comprises the following steps: S1: Based on the user's online activity data, text analysis is used to use big data analysis technology to identify user interests, extract the characteristics of interest points, and generate interest preference feature sets; S2: Based on the interest preference feature set, analyzing the user's interests, associating the user's behavior pattern with the interests, and generating an interest association map; S3: Based on the interest association graph, reconfigure the connection mode of nodes and edges, optimize the network structure, evaluate the paths between nodes of different types, and generate an optimized path set; S4: Based on the optimized path set, analyzing the information transmission efficiency, adjusting the weight distribution of paths and nodes, and generating an optimized user interest network; S5: Based on the optimized user interest network, use time series analysis and prediction of user interest trends, and combine historical behavior patterns to generate a behavior trend prediction model; S6: Based on the behavior trend prediction model, by predicting user behavior, accurately delivering advertisements that match user interests, and generating a targeted advertising content list; S7: Based on the targeted advertising content list, using big data processing technology, by monitoring advertising display, click-through rate and user browsing data, and adjusting the delivery strategy, generate advertising precision delivery effect analysis results.

[0018] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by using big data technology to analyze user browsing content and records, effectively extract keywords and identify points of interest, use content association analysis to deeply analyze the relationship between points of interest, accurately classify similar interest groups, accurately adjust the connection between nodes and edges in the network, optimize the information flow path, and further improve the accuracy of advertising content matching by predicting user behavior trends. Through the prediction model, screen and sort advertising content to ensure that each advertising delivery can achieve a high degree of personalization and accuracy. By using real-time monitoring of advertising effects and adjusting delivery strategies according to user feedback, advertising delivery efficiency and user satisfaction are significantly improved. Through in-depth analysis of user behavior and interests, accurate matching and delivery of advertising content is achieved, the pertinence and effectiveness of advertising are greatly enhanced, while optimizing user experience and avoiding interference from irrelevant advertising. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a system flow chart of the present invention; Figure 2 It is a schematic diagram of the system framework of the present invention; Figure 3 This is a flow chart of the user interest mining module of the present invention; Figure 4 It is a flow chart of the content association analysis module of the present invention; Figure 5 This is a flow chart of the heterogeneous network optimization module of the present invention; Figure 6 This is a flow chart of the user behavior insight module of the present invention; Figure 7 This is a flow chart of the advertisement matching strategy module of the present invention; Figure 8 This is a flow chart of the precise delivery effect monitoring module of the present invention; Fig. 9 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0022] Embodiment 1 See also Figure 1 to Figure 2 The present invention provides a technical solution, a media promotion directional and precise delivery system based on big data, including: The user interest mining module is based on user online behavior data and uses big data technology to analyze user browsing content and records, extract keywords, identify points of interest, classify positive and negative emotions, and generate interest preference feature sets; The content association analysis module analyzes the relationship between interest points, classifies similar interest groups, identifies interest trends, and generates interest association graphs based on interest preference feature sets; The heterogeneous network optimization module uses graph algorithms to adjust the connection mode of nodes and edges in the heterogeneous network based on the interest association graph, and uses the meta-path selection method to determine the paths between nodes of different types to generate an optimized user interest network. The user behavior insight module models user behavior sequences based on the optimized user interest network, analyzes the potential trends and motivations of the behaviors, and generates a behavior trend prediction model; The advertising matching strategy module is based on the behavioral trend prediction model, filters matching advertising content, sorts the user group suitability, and generates a targeted advertising content list; The precision delivery effect monitoring module monitors the effect of advertising based on the targeted advertising content list, adjusts the strategy according to user interaction, and generates analysis results of the precision delivery effect of advertising.

[0023] The interest preference feature set includes user topic preferences, active time periods and interaction frequencies. The interest association map includes the correlation between topics, interest evolution paths and group preference trends. The optimized user interest network includes node correlation areas, key propagation paths and information aggregation efficiency highlights. The behavior trend prediction model includes the probability of behavior occurrence, trend turning points and potential interest triggering factors. The targeted advertising content list includes the advertising topic suitability, target user matching and content innovation points. The results of the analysis of the effectiveness of precise advertising delivery include the click-through rate highlights, user satisfaction changes and delivery cost-effectiveness ratio.

[0024] See also Figure 2 and Figure 3 ,The user interest mining module includes: The content analysis submodule uses big data technology to analyze the user's browsing history based on the user's online behavior data, extracts keywords with increased frequency, identifies the user's points of interest through keywords, and generates the execution process of the point of interest identification result as follows; In the content analysis submodule, the browsing records in the user's online behavior data are analyzed through text mining technology and natural language processing methods. The bag-of-words model in the natural language processing method is used to convert the user's browsing text into a vector form for mathematical processing. The TF-IDF algorithm is used to calculate the weight of the vector to identify the high-importance words in each document, that is, keywords. It involves the calculation of the frequency of each word (TF) and the inverse frequency (IDF) of the word in all documents. The product result is the TF-IDF value, which is used to measure the importance of a word to the user's interest point, filter out keywords with increasing frequency, so as to accurately identify the user's interest point, and generate a detailed interest point identification result file, which records the changing trend of user interests in detail and provides basic data for subsequent modules.

[0025] The emotion classification submodule analyzes the user's comments and feedback on the interest points based on the interest point recognition results, distinguishes the emotional tendencies, and classifies the interest points and emotional tendencies to obtain the emotion classification results. The execution process is as follows; In the emotion classification submodule, the sentiment analysis algorithm is used to analyze the user's comments and feedback on multiple points of interest. The dictionary-based sentiment analysis method is used to analyze the text in the point of interest identification results sentence by sentence, identify the emotional words in the text, and score the emotional tendency of the words according to the sentiment dictionary, ranging from extremely negative to extremely positive. Through weighted calculation, the overall emotional tendency of each comment or feedback is determined, and the data is classified according to the point of interest and emotional tendency to generate the emotion classification result. It not only includes the user's emotional tendency for each point of interest, but also presents the pattern of user emotional distribution through statistical analysis, which is of great significance for understanding the user's emotional response and optimizing the user experience.

[0026] The feature set integration submodule associates and integrates the user's interest points with the emotional tendency based on the emotion classification results. By analyzing the user's interest and emotional feedback, the execution process of generating the interest preference feature set is as follows; In the feature set integration submodule, the user's interest points and emotional tendencies are comprehensively analyzed and integrated through data mining and machine learning technology. The Apriori algorithm in the association rule learning algorithm is used to analyze the correlation between each interest point and emotional tendency in the emotion classification results. This includes calculating key indicators such as the support, confidence, and lift of the item set to identify strong association rules. The rules reveal the intrinsic connection between the user's interest points and emotional tendencies, and generate a set of comprehensive interest preference feature sets. It not only reflects the user's interest points, but also integrates the user's emotional feedback, providing rich data support for subsequent personalized recommendations and content optimization.

[0027] See also Figure 2 and Figure 4 , the content association analysis module includes: The execution process of the interest point analysis submodule, based on the interest preference feature set, analyzes the user behavior data, identifies the user's interest points, extracts the key features of the interest points, and generates interest point keywords is as follows; In the POI analysis submodule, natural language processing (NLP) technology and text analysis algorithms are used to conduct in-depth analysis of user behavior data, including user browsing history and interactive content, remove irrelevant characters, perform part-of-speech tagging and stem extraction to normalize text data, and use the term frequency-inverse document frequency (TF-IDF) algorithm to calculate the importance of each word in the text, thereby identifying high-frequency keywords related to user POIs. Next, the Latent Dirichlet Allocation (LDA) topic model is used to further analyze keywords to extract topic keywords representing user POIs. The generated POI keyword file records the user's interest areas and preferences in detail, providing a basis for precision marketing and personalized recommendations.

[0028] The interest group classification submodule divides users into groups based on interest point keywords, and divides users into differentiated groups according to the similarity of their interest points. The execution process of generating the classification of similar interest groups is as follows; In the interest group classification submodule, users are divided into groups through a clustering algorithm. Based on the interest point keywords generated by the interest point analysis submodule, the K-means clustering algorithm is used to divide users into different groups according to the similarity of keywords. When executing the K-means algorithm, K center points are randomly selected, the distance between each user and the center point is calculated, and the users are classified to the nearest center point according to the distance to form a preliminary group. The position of the center point is iteratively adjusted until the group classification no longer changes. The generated classification file of similar interest groups reveals the group distribution of users based on interest points, which is of great value for understanding market segmentation and user group characteristics.

[0029] The trend identification submodule is based on the classification of similar interest groups and uses big data analysis technology to identify the development and change trends of interest points of multiple groups and generate interest association maps. The execution process is as follows; In the trend identification submodule, big data analysis technology is used to identify the trends of the classification results of similar interest groups. Time series analysis methods are used to combine the interest point change data of user groups. By calculating statistical indicators such as moving average and exponential smoothing, the development and change trends of different interest groups over time are identified. The association rule learning algorithm is used to analyze the correlation between different interest groups, find out the frequent co-occurring interest point combinations, and draw an interest association map. It not only shows the connection between interest groups, but also reveals the dynamic changes of interest points over time, which is of guiding significance for predicting future market trends and user interest changes.

[0030] See also Figure 2 and Figure 5 , the heterogeneous network optimization module includes: The network structure adjustment submodule is based on the interest association graph. If the connection between nodes does not meet the optimization goal, the connection mode of nodes and edges is reconfigured. The execution process of generating the adjusted network structure is as follows; In the network structure adjustment submodule, the connection mode of nodes and edges is reconfigured based on the interest association graph through graph theory and network optimization theory. The data format is nodes (representing users or interest points) and edges (representing the strength of association between interest points) in the interest association graph. Community detection algorithms, such as modularity optimization algorithms, are used to identify the community structure in the graph, that is, to gather closely connected nodes in the graph together to form communities. On this basis, the connections that do not meet the optimization goals are adjusted, and network flow algorithms, such as the maximum flow minimum cut algorithm, are used to determine the optimal connection mode between nodes, thereby reducing unnecessary edges and optimizing the global structure of the network. Considering the centrality indicators of nodes, such as degree centrality and betweenness centrality, to ensure that important nodes maintain high connectivity in the network, the generated adjusted network structure file records the new configuration of nodes and edges in detail, optimizes the information transmission efficiency of the network, and ensures the maximization of interest propagation.

[0031] The path selection submodule is based on the adjusted network structure. If there are multiple paths connecting nodes of different types, the meta-path selection method is used to determine the optimized path. The execution process of generating the optimized path set is as follows; In the path selection submodule, based on the adjusted network structure, the meta-path selection method is used to determine the optimized path. By analyzing the nodes and edges in the network structure, the meta-path selection method defines the composite relationship between different types of nodes and edges to find the best path connecting differentiated types of nodes. Use graph-based query languages, such as Cypher, to express meta-path queries to identify a set of paths that meet a specific pattern. Based on the length of the path, the weight of the edge, and the attributes of the nodes on the path, a dynamic programming algorithm is used to calculate the comprehensive score of each path, so that the path with the highest score is selected as the optimized path. The generated optimized path collection file records the best path connecting different types of nodes. The optimized path provides a basis for the efficient dissemination of information in the network.

[0032] The network efficiency optimization submodule is based on the optimization path set. If the information transmission efficiency in the network is delayed, the weight distribution of the paths and nodes is adjusted to generate the optimized user interest network. The execution process is as follows; In the network efficiency optimization submodule, based on the optimized path set, the network optimization theory and algorithm are used to adjust the weight distribution of paths and nodes, analyze the path information in the optimized path set, including the nodes and edges on the path, and use the linear programming method to model the information transmission efficiency in the network. The goal is to minimize the total delay of information transmission. In this model, the weight of each edge and the processing speed of each node are adjusted to optimize the information transmission path. The adjustment method includes increasing the weight of efficient paths, reducing the frequency of use of inefficient paths, and dynamically adjusting the processing capacity of nodes according to network traffic. The generated optimized user interest network file shows an efficient information transmission network, the information transmission efficiency is significantly improved, the delay is minimized, and the user interest information is transmitted quickly and accurately.

[0033] See also Figure 2 and Figure 6 , the user behavior insight module includes: The behavior sequence modeling submodule collects user behavior sequence data based on the optimized user interest network. If the user activity presents a periodic pattern, it is marked as a periodic behavior. The user's behavior sequence is modeled. The execution process of generating the user behavior model is as follows; In the behavior sequence modeling submodule, the user behavior sequence data collected in the optimized user interest network, including the user's behavior log, is processed and analyzed through time series analysis and periodicity detection algorithms to record the user's activity information at different time points. A convolutional neural network model is applied to capture the time dependency in the behavior sequence, while Fourier transform analysis is combined to identify the periodic patterns in the sequence. The LSTM model can be used to understand and simulate the complex patterns of user behavior over time, while Fourier transform analysis helps to identify whether the behavior is periodic, such as patterns that repeat weekly or monthly. If a periodic pattern is found, the behavior is marked as periodic in the model. The generated user behavior model not only depicts the time series characteristics of user behavior in detail, but also identifies the periodic patterns of behavior, providing a basis for subsequent trend analysis and motivation identification.

[0034] The trend analysis submodule is based on the user behavior model and applies convolutional neural networks to analyze the time series of behavior data. If the behavior trend increases, it is marked as a growth trend, and the potential trend of the behavior is identified. The execution process of generating the potential trend analysis results is as follows; In the trend analysis submodule, based on the user behavior model, the convolutional neural network (CNN) is applied to analyze the time series of behavior data, collect and process user behavior data, such as click volume, browsing time, etc., to form a normalized input matrix. The convolutional neural network extracts features from the time series data through multiple convolutional layers and pooling layers. Each convolutional layer convolves the data through a set of learned filters to extract local features; the pooling layer is responsible for reducing the spatial dimension of the features and enhancing the generalization ability of the model. Nonlinear processing is added through the ReLU activation function to help the model capture more complex data patterns. After continuous convolution and pooling operations, the model can extract key features from the original time series and conduct a comprehensive analysis of the features through the fully connected layer to predict behavioral trends. In the behavioral trend prediction stage, the softmax function is used to output the probabilities of different trend categories. If the trend of a certain behavior increases, it is marked as a growth trend through the threshold judgment. The CNN parameters are continuously adjusted through the back propagation and gradient descent algorithms to optimize the model performance. The generated potential trend analysis results are output in the form of reports or labels, clearly indicating which user behaviors show a growth trend, providing data support for subsequent decision-making. By accurately identifying potential trends in behavior, we can effectively guide marketing strategy adjustments and product optimization, enhance user experience and improve corporate benefits.

[0035] The convolutional neural network algorithm follows the formula:

[0036]

[0037]

[0038] in, is the output of the convolutional layer, is the original weight of the convolution kernel, is the activation function, is the dynamic adjustment of weight, is the weight adjustment coefficient, The reference time window The input data, is the original bias term, is the dynamic adjustment of the bias term, is the bias adjustment coefficient, is the original weight of the fully connected layer, is the dynamic adjustment amount of the fully connected layer weight, is the weight adjustment coefficient of the fully connected layer, is the original bias term of the fully connected layer, is the dynamic adjustment of the bias term of the fully connected layer, is the bias adjustment coefficient of the fully connected layer, is the output of the fully connected layer, is the activation function used for the fully connected layer, is the result of trend analysis. is an improved trend identification function, and New parameters introduced for trend analysis; Execution process: The weights and biases of the convolutional layer and the fully connected layer are dynamically adjusted. , , and Coefficients and corresponding adjustments , , and , to achieve real-time optimization of weights and biases, using training results or real-time feedback, Consider time windows , so that the model can better adapt to the changes of data at different time scales, introducing new parameters and , the statistical characteristics of network output and advanced mathematical models are used to enhance the accuracy of trend judgment. Innovative parameter introduction and adjustment strategies are designed to improve the sensitivity and accuracy of CNN models in trend analysis.

[0039] The motivation identification submodule analyzes the motivation behind the user behavior based on the potential trend analysis results. If the target behavior is related to the user's interest point, it is identified as interest-driven behavior. The user's behavior motivation is identified from the trend and behavior pattern, and the execution process of generating a behavior trend prediction model is as follows; In the motivation identification submodule, based on the results of potential trend analysis, the motivation behind user behavior is further analyzed. By combining the data of points of interest in the user interest network, association rule mining and causal reasoning models are used to analyze the correlation between user behavior and points of interest, aiming to identify behaviors driven by user interests and to mine user behavior motivations from trends and behavior patterns. Association rule mining reveals the strong correlation between user behavior and specific points of interest, while causal reasoning models help understand the motivations behind the correlations. The generated behavior trend prediction model not only captures the trends and patterns of user behavior, but also deeply analyzes the motivations behind the trends and patterns, providing motivational insights for understanding user behavior, thereby supporting more accurate user behavior predictions and personalized recommendations.

[0040] See also Figure 2 and Figure 7 , the advertisement matching strategy module includes: The advertisement content screening submodule analyzes user behavior data based on the behavior trend prediction model, identifies advertisement content associated with user interests, selects advertisements that match the user behavior pattern, and generates a matching advertisement set. The execution process is as follows: In the advertising content screening submodule, the advertising content associated with the user's interests is identified through machine learning classification algorithms and user behavior data analysis. The behavioral trend prediction model is used to analyze the user's behavior log data, including behavioral indicators such as the user's browsing history, click-through rate, and dwell time. The decision tree algorithm is used to build a classification model based on the relationship between user behavior patterns and advertising features to score and screen advertising content. The algorithm comprehensively considers the user's historical behavior data and interest preferences, automatically identifies advertising content that is highly matched with the user's behavior pattern, and marks the content. The generated matching advertising collection file records the screened advertising content in detail. The content is highly correlated with the user's interests and behavior patterns, providing accurate data support for personalized advertising recommendations.

[0041] The suitability ranking submodule evaluates the suitability of the advertisement content based on the matching advertisement set, and sorts the advertisement content according to the suitability with reference to the diversity of user interests and the frequency of behaviors. The execution process of generating the sorted advertisement list is as follows; In the fitness ranking submodule, based on the matching advertisement set, the advertisement content is ranked by the fitness evaluation algorithm. The collaborative filtering and content analysis techniques are used to analyze the advertisement content in the matching advertisement set, involving the click-through rate of the advertisement, user interaction data, and the relevance to the user's interest points. According to the diversity of user interests and the frequency of behavior, the weighted ranking algorithm is used to calculate the fitness score of each advertisement. The advertisements are ranked according to the fitness score based on the relevance of the advertisement content, the user's feedback on the advertisement, and the breadth of the user's interests. The generated ranked advertisement list shows the ranking of each advertisement content in detail, ensuring that the advertisements seen by the user are highly consistent with their interests and behavior patterns, thereby improving the click-through rate of the advertisement and user satisfaction.

[0042] The content list generation submodule uses big data analysis technology to filter advertising content associated with user interests and behavior trends based on the sorted advertising list, and generates a targeted advertising content list. The execution process is as follows; In the content list generation submodule, based on the sorted advertising list, big data analysis technology is used to screen advertising content associated with user interests and behavioral trends, and user behavior data analysis and prediction models are used in combination, including user browsing behavior, social media interaction, and purchase history. Through complex event processing (CEP) and association rule learning algorithms, user behavior data and advertising characteristics are deeply analyzed to identify advertising content that is most relevant to user interests and behavioral trends. The algorithm conducts detailed screening based on the matching degree between the target effect of the advertisement and user needs, and gives priority to advertising content that can arouse user interest or meet user needs. The generated targeted advertising content list accurately reflects the user's interest preferences and behavioral trends, providing users with highly relevant and valuable advertising content, thereby improving the effectiveness of advertising and user experience.

[0043] See also Figure 2 and Figure 8 , the precise delivery effect monitoring module includes: The advertising performance monitoring submodule collects the number of ad impressions, click-through rates, and initial user responses based on the targeted ad content list, establishes performance indicators for the ads, and generates the following execution flow for ad performance monitoring data; In the advertising performance monitoring submodule, data collection and analysis tools are used to collect data such as the number of impressions, click-through rates, and initial user reactions of each advertisement in the targeted advertising content list, covering key performance indicators such as advertising ID, number of impressions, number of clicks, and user reaction time. Using statistical analysis methods, the collected data is summarized and processed to calculate the average number of impressions, total click-through rate, and average user reaction speed of each advertisement, and to establish advertising performance indicators. Using algorithms such as A / B testing and time series analysis, the trend of advertising performance changes over time is further analyzed to identify which advertising content performs best in a specific time period. The generated advertising performance monitoring data file records the performance evaluation results of each advertisement in detail, providing an accurate data basis for subsequent conversion rate analysis and delivery strategy optimization.

[0044] The conversion rate analysis submodule analyzes the user's purchase behavior after clicking on an ad or completing a form based on the ad performance monitoring data, identifies the conversion rate of multiple ad contents, and generates the conversion rate analysis results. The execution process is as follows: In the conversion rate analysis submodule, based on the advertising performance monitoring data, through conversion tracking and user behavior analysis technology, the behavior patterns of users after clicking on the advertisement are deeply analyzed, including purchase behavior or completing form filling, including user identification, advertisement click time, behavior completion type and time, etc. The conversion funnel analysis method is adopted to analyze the conversion path of users from clicking on the advertisement to completing the target behavior, and the conversion rate of each stage is calculated to identify efficient and inefficient advertising content. Regression analysis and machine learning models such as logistic regression are used to further explore the influencing factors behind the conversion rate of different advertising content and generate a conversion rate analysis result file. The results reveal the conversion efficiency of multiple advertising content, which is of great value for understanding user behavior and optimizing advertising content.

[0045] The delivery strategy optimization submodule evaluates the effects of multiple advertising contents based on the conversion rate analysis results, proposes adjustment plans for inefficient advertisements, optimizes the target audience settings for precise advertising delivery, and obtains the execution process of the analysis results of precise advertising delivery effects as follows; In the delivery strategy optimization submodule, based on the conversion rate analysis results, market analysis and user profiling technology are used to evaluate the overall effect of multiple advertising contents, and adjustment plans are proposed for poorly performing advertising contents. A multi-dimensional evaluation method is used to comprehensively evaluate the delivery effect of each advertisement, combining the conversion rate of the advertisement, the feedback of the target audience and market trends. For advertisements with low conversion rates, decision tree analysis and cluster analysis methods are used to identify the reasons for inefficiency, such as inaccurate target audiences and unattractive advertising creativity, and adjust the target audience settings or content creativity of the advertisement accordingly. Through simulation tests and predictive analysis, the advertising delivery strategy is optimized and a file of analysis results of the precise advertising delivery effect is generated. The expected effect of the optimized advertisement is displayed to guide advertisers to achieve more accurate advertising delivery and improve advertising effect and return on investment.

[0046] The method for targeted and precise delivery of media promotion based on big data is implemented based on the above-mentioned targeted and precise delivery system for media promotion based on big data, and includes the following steps: S1: Based on the user's online activity data, text analysis is used to use big data analysis technology to identify user interests, extract the characteristics of interest points, and generate interest preference feature sets; S2: Based on the interest preference feature set, analyze the user's interests, associate the user's behavior pattern with the interests, and generate an interest association map; S3: Based on the interest association graph, reconfigure the connection mode of nodes and edges, optimize the network structure, evaluate the paths between nodes of different types, and generate an optimized path set; S4: Based on the optimized path set, analyze the information transmission efficiency, adjust the weight distribution of paths and nodes, and generate the optimized user interest network; S5: Based on the optimized user interest network, use time series analysis and prediction of user interest trends, combined with historical behavior patterns, to generate a behavior trend prediction model; S6: Based on the behavior trend prediction model, by predicting user behavior, accurately deliver advertisements that match user interests and generate a targeted advertising content list; S7: Based on the targeted advertising content list, using big data processing technology, by monitoring advertising display, click-through rate and user browsing data, and adjusting the delivery strategy, generate advertising precision delivery effect analysis results.

[0047] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A media promotion directional and precise delivery system based on big data, characterized in that: The system comprises: The user interest mining module is based on user online behavior data and uses big data technology to analyze user browsing content and records, extract keywords, identify points of interest, classify positive and negative emotions, and generate interest preference feature sets; The content association analysis module analyzes the interest point relationship based on the interest preference feature set, classifies similar interest groups, identifies interest trends, and generates an interest association map; The heterogeneous network optimization module uses a graph algorithm to adjust the connection mode of nodes and edges in the heterogeneous network based on the interest association graph, uses a meta-path selection method to determine the path between nodes of different types, and generates an optimized user interest network; The user behavior insight module models the user behavior sequence based on the optimized user interest network, analyzes the potential trends and motivations of the behavior, and generates a behavior trend prediction model; The advertisement matching strategy module screens matching advertisement contents, sorts user group suitability, and generates a targeted advertisement content list based on the behavior trend prediction model; The precision delivery effect monitoring module monitors the advertising delivery effect based on the targeted advertising content list, adjusts the strategy according to user interaction, and generates the advertising precision delivery effect analysis result; Among them, the interest preference feature set includes user topic preferences, active time periods and interaction frequencies, the interest association map includes the correlation between topics, interest evolution paths and group preference trends, the optimized user interest network includes node correlation areas, key propagation paths and information aggregation efficiency highlights, the behavior trend prediction model includes behavior occurrence probability, trend turning points and potential interest triggering factors, the targeted advertising content list includes advertising topic suitability, target user matching and content innovation points, and the advertising precision delivery effect analysis results include click-through rate highlights, user satisfaction changes and delivery cost-effectiveness ratio.

2. According to the big data-based media promotion directional and precise delivery system according to claim 1, it is characterized in that: The user interest mining module includes: The content analysis submodule uses big data technology to analyze users' browsing history based on their online behavior data, extract keywords with increasing frequency, identify users' points of interest through keywords, and generate points of interest identification results; The emotion classification submodule analyzes the user's comments and feedback on the interest points based on the interest point recognition results, distinguishes the emotional tendencies, and classifies the interest points and emotional tendencies to obtain the emotion classification results; The feature set integration submodule associates and integrates the user's interest points with the emotional tendency based on the emotion classification result, and generates an interest preference feature set by analyzing the user's interest and emotional feedback.

3. The media promotion directional and precise delivery system based on big data according to claim 2 is characterized in that: The content association analysis module includes: The interest point analysis submodule analyzes the user behavior data based on the interest preference feature set, identifies the user's interest points, extracts the key features of the interest points, and generates interest point keywords; The interest group classification submodule divides users into groups based on the interest point keywords, divides users into differentiated groups according to the similarity of interest points, and generates similar interest group classifications; The trend identification submodule is based on the classification of the same interest groups and uses big data analysis technology to identify the development and change trends of the interest points of multiple groups and generate an interest association map.

4. The media promotion directional and precise delivery system based on big data according to claim 3 is characterized in that: The heterogeneous network optimization module includes: The network structure adjustment submodule is based on the interest association graph. If the connection between nodes does not meet the optimization target, the connection mode of nodes and edges is reconfigured to generate an adjusted network structure. The path selection submodule is based on the adjusted network structure, and if there are multiple paths connecting nodes of different types, a meta-path selection method is used to determine the optimized path and generate an optimized path set; The network efficiency optimization submodule is based on the optimized path set. If the information transmission efficiency in the network is delayed, the optimized user interest network is generated by adjusting the weight distribution of the paths and nodes.

5. The media promotion directional and precise delivery system based on big data according to claim 4 is characterized in that: The user behavior insight module includes: The behavior sequence modeling submodule collects user behavior sequence data based on the optimized user interest network. If the user activity presents a periodic pattern, it is marked as a periodic behavior, and the user's behavior sequence is modeled to generate a user behavior model. The trend analysis submodule applies a convolutional neural network based on the user behavior model to analyze the time series of the behavior data. If the behavior trend increases, it is marked as a growth trend, the potential trend of the behavior is identified, and a potential trend analysis result is generated; The convolutional neural network algorithm follows the formula: ； ； ； in, is the output of the convolutional layer, is the original weight of the convolution kernel, is the activation function, is the dynamic adjustment of weight, is the weight adjustment coefficient, The reference time window The input data, is the original bias term, is the dynamic adjustment of the bias term, is the bias adjustment coefficient, is the original weight of the fully connected layer, is the dynamic adjustment amount of the fully connected layer weight, is the weight adjustment coefficient of the fully connected layer, is the original bias term of the fully connected layer, is the dynamic adjustment of the bias term of the fully connected layer, is the bias adjustment coefficient of the fully connected layer, is the output of the fully connected layer, is the activation function used for the fully connected layer, is the result of trend analysis. is an improved trend identification function, and New parameters introduced for trend analysis; The motivation identification submodule analyzes the motivation behind the user behavior based on the potential trend analysis results. If the target behavior is related to the user's interest point, it is identified as interest-driven behavior. The user's behavioral motivation is identified from trends and behavioral patterns, and a behavioral trend prediction model is generated.

6. The media promotion directional and precise delivery system based on big data according to claim 5 is characterized in that: The advertisement matching strategy module includes: The advertisement content screening submodule analyzes the user behavior data based on the behavior trend prediction model, identifies the advertisement content associated with the user's interests, selects advertisements that match the user's behavior pattern, and generates a matching advertisement set; The suitability ranking submodule evaluates the suitability of the advertisement content based on the matching advertisement set, ranks the advertisement content according to the suitability with reference to the diversity of user interests and the frequency of behaviors, and generates a ranked advertisement list; The content list generation submodule uses big data analysis technology to screen advertising content associated with user interests and behavior trends based on the sorted advertising list and generates a targeted advertising content list.

7. The media promotion directional and precise delivery system based on big data according to claim 6 is characterized in that: The precise delivery effect monitoring module includes: The advertising performance monitoring submodule collects the number of advertisement displays, click-through rates, and initialization responses of users based on the targeted advertisement content list, establishes performance indicators for the advertisements, and generates advertising performance monitoring data; The conversion rate analysis submodule analyzes the user's purchase behavior or form filling after clicking on the advertisement based on the advertisement performance monitoring data, identifies the conversion rate of multiple advertisement contents, and generates a conversion rate analysis result; The delivery strategy optimization submodule evaluates the effects of multiple advertising contents based on the conversion rate analysis results, proposes adjustment plans for inefficient advertisements, optimizes the target audience settings for precise advertising delivery, and obtains analysis results of the precise advertising delivery effects.

8. A method for targeted and precise delivery of media promotion based on big data, characterized in that: The implementation of the media promotion directional and precise delivery system based on big data according to any one of claims 1 to 7 comprises the following steps: Based on the user's online activity data, text analysis and big data analysis technology are used to identify the user's points of interest, extract the features of the points of interest, and generate interest preference feature sets; Based on the interest preference feature set, analyzing the user's interests, associating the user's behavior pattern with the interests, and generating an interest association map; Based on the interest association graph, reconfigure the connection mode of nodes and edges, optimize the network structure, evaluate the paths between nodes of different types, and generate an optimized path set; Based on the optimized path set, analyzing the information transmission efficiency, adjusting the weight distribution of paths and nodes, and generating an optimized user interest network; Based on the optimized user interest network, using time series analysis and prediction of user interest trends, combined with historical behavior patterns, to generate a behavior trend prediction model; Based on the behavior trend prediction model, by predicting user behavior, accurately delivering advertisements that match user interests, and generating a targeted advertising content list; Based on the targeted advertising content list, big data processing technology is used to monitor advertising display, click-through rate and user browsing data, and adjust the delivery strategy to generate analysis results of the precise advertising delivery effect.

Images