Electronic coupon recommendation obtaining method and system based on social network behaviors
By acquiring user behavior data from social networks, calculating social influence and interest preference graphs, and constructing a deep neural network model, personalized electronic coupon recommendations are provided to new users. This solves the problem of insufficient new user recommendations in traditional methods, achieves accurate recommendations and viral marketing, and protects user privacy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUZHOU HUIFA INFORMATION TECH CO LTD
- Filing Date
- 2026-01-12
- Publication Date
- 2026-04-10
AI Technical Summary
Existing technologies struggle to provide effective electronic coupon recommendations for new or low-frequency consumers, and fail to leverage users' social relationships on social networks for viral spread. Traditional methods lack the integration of multi-dimensional, fine-grained social behaviors and preferences.
By acquiring multi-dimensional behavioral data of target users and their socially connected users, we calculate the social influence index and dynamic interest preference map, construct a deep neural network model based on the attention mechanism, combine it with the collaborative filtering algorithm to generate personalized electronic coupon recommendations, and push and iteratively optimize them through social network platforms.
It enables precise recommendations for new users, leverages social influence for viral marketing, dynamically captures changes in user interests, improves the accuracy of recommendations and marketing effectiveness, while protecting user privacy and complying with data regulatory requirements.
Smart Images
Figure CN121836797A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of computer, more particularly, to an electronic coupon recommendation acquisition method and system based on social network behavior. BACKGROUND
[0002] With the development of e-commerce and mobile payment, electronic coupons have become an important marketing tool for merchants to attract, promote and improve sales. Traditional electronic coupon recommendation methods mainly rely on users' own consumption history (such as purchase records, browsing records), and predict the goods or coupons that users may be interested in the future by analyzing their past purchase records.
[0003] However, such methods have obvious limitations: first, for new users or low-frequency consumption users, transaction data is insufficient, making it difficult to generate effective recommendations; second, only reflecting the completed consumption behavior of users, it is impossible to capture their potential, unconverted purchase interests; in addition, the role and influence of users in the social network are not considered, missing the opportunity to use their social relationships for propagation fission.
[0004] In recent years, some research has attempted to introduce social relationship information, but is limited to direct friend relationships or simple interaction frequency, failing to deeply integrate multi-dimensional, fine-grained social behavior and preferences and influence derived from behavior.
[0005] Therefore, there is an urgent need for an electronic coupon intelligent recommendation technical solution that can deeply integrate user social network behavior analysis and achieve more accurate and influential recommendations. SUMMARY
[0006] The purpose of the present application is to solve the problems existing in the prior art, and to provide an electronic coupon recommendation acquisition method and system based on social network behavior.
[0007] To solve the above problems, the present application adopts the following technical solution:
[0008] The electronic coupon recommendation acquisition method based on social network behavior comprises the following steps:
[0009] S1: Obtain multi-dimensional behavior data of a target user and his / her social association users from a social network platform, the behavior data including consumption preference data, social interaction data and content generation data of the users;
[0010] S2: Clean and structure the multi-dimensional behavior data, and extract a user social behavior feature vector;
[0011] S3: Calculate the social influence index and dynamic interest preference graph of the target user based on the user social behavior feature vector;
[0012] S4: Obtain a candidate electronic coupon set and a label attribute of each electronic coupon, the label attribute being at least associated with a product category, an applicable scenario, a merchant brand, and a promotion intensity;
[0013] S5: Build a recommendation model, perform multi-modal matching on a social influence index of a target user, a dynamic interest preference graph, and a label attribute of a candidate electronic coupon, introduce a collaborative filtering algorithm, fuse historical coupon redemption data of other users having similar social behavior characteristics with the target user, and calculate and generate a personalized electronic coupon recommendation list for the target user;
[0014] S6: Push the personalized electronic coupon recommendation list to the target user through a message interface or an associated application interface of a social network platform, and record feedback behavior data of the user;
[0015] S7: Iteratively optimize the recommendation model according to the feedback behavior data.
[0016] As a further scheme of the application: in step S1, the consumption preference data is obtained through analysis of historical transaction records, product browsing records, and collection behaviors of the user;
[0017] The social interaction data includes frequencies and objects of likes, comments, and forwarding of the user on coupon-related content.
[0018] The content generation data includes text and image content related to consumption experience published by the user.
[0019] As a further scheme of the application: in step S3, the social influence index is quantified through a network centrality index of a user node, content propagation breadth, and an interaction weight.
[0020] The dynamic interest preference graph is dynamically generated through theme extraction and sentiment analysis on content published and interacted by the user by using a natural language processing technology, and in combination with a time decay factor of behavior occurrence.
[0021] As a further scheme of the application: the network centrality index includes at least one of degree centrality, betweenness centrality, and eigenvector centrality.
[0022] The interaction weight is differentiated and assigned according to self-influence of an interactive object and an interaction type.
[0023] As a further scheme of the application: in step S5, the multi-modal matching adopts a deep neural network model based on an attention mechanism, the model having an input of embedded representations of the user social behavior feature vector and the coupon label attribute, and an output of matching scores.
[0024] As a further scheme of the present application, in step S6, the feedback behavior data includes clicking, collecting, sharing, and canceling behavior of the recommended electronic coupon.
[0025] The present application provides a system for implementing the above method, comprising:
[0026] A data collection module is configured to collect multi-dimensional social behavior data of a target user through a social network platform in a secure manner; the module preferably comprises a privacy management unit configured to verify the authorization status of the user before collecting data and to anonymize and desensitize the collected personal information.
[0027] A data processing and analysis module is configured to clean, structure, and extract features from the social behavior data, and to calculate the social influence index and dynamic interest preference graph of the user.
[0028] A coupon management module is configured to maintain a candidate electronic coupon set and its label attributes.
[0029] An intelligent recommendation module is configured to generate a personalized electronic coupon recommendation list by fusing the analysis results of the user's social behavior, coupon attributes, and collaborative filtering signals.
[0030] A push and feedback module is configured to push the recommendation list to the target user and collect feedback behavior data of the user.
[0031] A model optimization module is configured to continuously train and update the recommendation model based on the feedback behavior data.
[0032] As a further scheme of the present application, the data processing and analysis module comprises:
[0033] A feature extraction unit is configured to convert unstructured social behavior data into a feature vector.
[0034] An influence calculation unit is configured to quantify the social influence of the user based on a graph calculation algorithm.
[0035] An interest modeling unit is configured to construct a dynamic interest preference graph based on natural language processing and time series analysis.
[0036] The present application also provides a computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.
[0037] Compared with the prior art, the present application has the following advantages:
[0038] First, the present application fuses multi-source data such as consumption preferences, social interactions and original content, and the dynamic interest preference graph constructed can more deeply and more forwardly understand user interests, effectively alleviate the data sparsity and cold start problems, and the recommendation results are more accurate and can bring surprise.
[0039] Second, the present application can identify key opinion leaders or high transmission potential users by quantifying social influence indexes, recommend more attractive or sharing incentive attribute coupons for them, thereby realizing secondary transmission and fission marketing of the coupons by using their social networks, and amplifying marketing effects.
[0040] Third, the interest graph of the user of the present application is dynamically updated, can timely capture user interest migration, and makes the recommendation results keep relevant with time changes, overcoming the lag of the static historical data model.
[0041] Fourth, the present application organically fuses various technologies such as social network analysis, natural language processing, graph computing and deep learning recommendation model, realizes multi-modal and deep understanding and matching of complex social behaviors, and is advanced in technology and strong in self-adaptability.
[0042] Fifth, by introducing compliance checking and anonymous desensitization mechanism at the data collection end, the recommendation effect is ensured, the respect for user privacy is embodied, the current strict data supervision requirements are met, the compliance risk is reduced, and the user trust is enhanced. BRIEF DESCRIPTION OF DRAWINGS
[0043] Figure 1 is the overall flowchart of the method of the present application;
[0044] Figure 2 is the module structure schematic diagram of the system of the present application. DETAILED DESCRIPTION
[0045] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application; obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments, and all other embodiments obtained by those skilled in the art without creative labor based on the embodiments in the present application belong to the protection scope of the present application.
[0046] Please refer to Figure 1 , the electronic coupon recommendation acquisition method based on social network behavior includes the following steps:
[0047] S1, safe data collection:
[0048] With the explicit authorization of the user, obtain multi-dimensional behavior data of the target user and his social association users (such as friends, mutual followers) from social network platforms (such as WeChat, Weibo, Xiaohongshu, etc.) and secure data interfaces. Behavior data specifically includes:
[0049] Consumer preference data: Through the associated e-commerce platform or payment tool (additional authorization required), analyze the user's historical transaction records (purchase product category, price), product detail page browsing duration and frequency, product collection and purchase behavior.
[0050] Social interaction data: User likes, comments (including emotional tendencies), and forwarding behavior and their frequency on platform-related content (including official accounts, other user shares) related to goods, brands, and promotional activities, as well as the ID of the interaction object.
[0051] Content generation data: User-initiated original posts, notes, videos, and other content, especially text and image / video information containing consumer experience, product evaluation, and shopping sharing.
[0052] After collection, immediately anonymize and replace information that can directly identify personal identity (such as user ID, mobile phone number), and desensitize sensitive information in text content.
[0053] S2, data processing and feature extraction:
[0054] Clean (de-duplicate, correct errors, fill in missing values) and standardize the collected multi-source heterogeneous data. Use natural language processing tools to vectorize text content; use image recognition technology to extract object and scene features from image content. Abstract users and their social association users as graph nodes, and abstract attention and interaction relationships as edges to construct a user social relationship graph. Finally, all information is integrated to extract a high-dimensional user social behavior feature vector.
[0055] S3, calculate social influence index and dynamic interest preference graph:
[0056] Based on the user social behavior feature vector, calculate the target user's social influence index and dynamic interest preference graph, as follows:
[0057] Social influence index calculation: On the social relationship graph constructed in step S2, calculate multiple network centrality indicators (such as degree centrality, eigenvector centrality) of the user. At the same time, quantify the propagation breadth (forwarding level) and depth (number of comments) of the original content. In addition, different types of interactions (such as "forward" is more important than "like") and interactions with different influence objects are weighted. Comprehensive the above factors, a standardized social influence index is calculated through a pre-set weighting formula.
[0058] Dynamic interest preference graph construction: topic extraction and sentiment analysis are performed on the text content published and interacted by the user to identify the active interest topics. An initial weight is assigned to each interest topic, and a time decay function is introduced to increase the weight of recently active topics and reduce the weight of topics that have not been involved for a long time. Finally, a vector with interest topics as dimensions and weights dynamically changing over time is formed, i.e. a dynamic interest preference graph.
[0059] S4, coupon management:
[0060] Obtain a set of candidate electronic coupons and the label attributes of each electronic coupon, and the label attributes are at least associated with product categories, applicable scenarios, merchant brands, and promotion intensity;
[0061] S5, building a recommendation model and generating a recommendation list:
[0062] Build a recommendation model, and perform multi-modal matching of the social influence index of the target user, the dynamic interest preference graph, and the label attributes of the candidate electronic coupons. The multi-modal matching adopts a deep neural network model based on an attention mechanism, and the input of the model is an embedding table of the user social behavior feature vector and the coupon label attributes.
[0063] The model automatically learns the importance of the association between different interest dimensions of the user and different coupon attributes through the attention mechanism. At the same time, the system finds the historical coupon redemption data of other users with similar social behavior characteristics as the target user as positive sample signals, and inputs them into the model for collaborative filtering learning.
[0064] The model calculates a final matching score for each candidate coupon, and generates a personalized electronic coupon recommendation list for the target user according to the score ranking.
[0065] S6, personalized push and feedback loop:
[0066] The personalized electronic coupon recommendation list is pushed to the target user through the message interface or the associated application program interface of the social network platform, and the push text can be personalized. The system tracks and records the feedback behavior data of the user, such as clicking, taking, collecting, sharing, and redeeming the recommended electronic coupons.
[0067] S7, continuous optimization of the model:
[0068] According to the feedback behavior data as new training samples, the recommendation model is iteratively optimized, "taking" and "redeeming" are regarded as strong positive feedback, and "ignoring" is regarded as negative feedback, which is used for incremental training or periodic retraining of the recommendation model, and the model parameters are continuously adjusted to make the recommendation effect continuously optimized with data accumulation.
[0069] Please refer toFigure 2 The embodiment provides a system for implementing the method.
[0070] The data collection module is used for collecting multi-dimensional social behavior data of a target user through a social network platform safely. The module comprises a privacy management unit, which is used for verifying an authorization state of a user before collecting data and performing anonymization and desensitization processing on collected personal information.
[0071] The data processing and analysis module is used for cleaning, structuring and feature extraction of the social behavior data, and calculating a social influence index and a dynamic interest preference graph of the user. The module comprises:
[0072] The feature extraction unit is used for converting unstructured social behavior data into a feature vector;
[0073] The influence calculation unit is used for quantifying social influence of the user based on a graph calculation algorithm;
[0074] The interest modeling unit is used for constructing a dynamic interest preference graph based on natural language processing and time series analysis.
[0075] The coupon management module is used for maintaining a candidate electronic coupon set and label attributes thereof;
[0076] The intelligent recommendation module is used for fusing a user social behavior analysis result, coupon attributes and collaborative filtering signals, and generating a personalized electronic coupon recommendation list;
[0077] The push and feedback module is used for pushing the recommendation list to the target user and collecting feedback behavior data of the user;
[0078] The model optimization module is used for continuously training and updating the recommendation model according to the feedback behavior data.
[0079] The embodiment also provides a computing device, which comprises at least one processor and a memory. The memory stores a computer program, and when the program is executed by the processor, the computing device can complete each step of the above-mentioned electronic coupon recommendation acquisition method based on social network behavior. The computing device can be a cloud server, a server cluster or an edge computing node.
[0080] The above merely describes a preferred embodiment of the present application; however, the protection scope of the present application is not limited to this. Any person skilled in the art can make equivalent replacements or changes to the technical scheme of the present application and the improved concept thereof within the technical scope disclosed by the present application, which should be covered by the protection scope of the present application.
Claims
1. A method for obtaining an electronic coupon recommendation based on social network behavior, characterized in that, The method comprises the following steps: S1: obtaining multi-dimensional behavior data of a target user and his / her social association users from a social network platform, wherein the behavior data comprises consumption preference data, social interaction data and content generation data of the user; S2: cleaning and structuring the multi-dimensional behavior data, and extracting a user social behavior feature vector; S3: calculating a social influence index and a dynamic interest preference graph of the target user based on the user social behavior feature vector; S4: obtaining a candidate electronic coupon set and label attributes of each electronic coupon, wherein the label attributes are at least associated with product categories, applicable scenarios, merchant brands and promotion intensities; S5: constructing a recommendation model, performing multi-modal matching between the social influence index and the dynamic interest preference graph of the target user and the label attributes of the candidate electronic coupons, introducing a collaborative filtering algorithm, fusing historical coupon redemption data of other users with similar social behavior characteristics to the target user, and calculating a personalized electronic coupon recommendation list for the target user; S6: pushing the personalized electronic coupon recommendation list to the target user through a message interface or an associated application interface of the social network platform, and recording feedback behavior data of the user; S7: iteratively optimizing the recommendation model according to the feedback behavior data. 2.The social network behavior based electronic coupon recommendation acquisition method of claim 1, wherein, In step S1, the consumption preference data is obtained by analyzing historical transaction records, product browsing records and collection behaviors of the user; The social interaction data comprises the frequency and objects of the user's likes, comments and forwards on coupon-related content; The content generation data comprises text and image content related to consumption experience published by the user. 3.The social network behavior based electronic coupon recommendation acquisition method of claim 1, wherein, In step S3, the social influence index is quantified by a network centrality index, content propagation breadth and interaction weight of the user node; The dynamic interest preference graph is dynamically generated by subject extraction and sentiment analysis on the content published and interacted by the user through natural language processing technology, and combined with a time decay factor of behavior occurrence. 4.The social network behavior based electronic coupon recommendation acquisition method of claim 3, wherein, The network centrality index comprises at least one of degree centrality, betweenness centrality and eigenvector centrality; The interaction weight is differentiated according to the influence of the interaction object and the interaction type. 5.The social network behavior based electronic coupon recommendation acquisition method of claim 1, wherein, In step S5, the multi-modal matching adopts a deep neural network model based on an attention mechanism, wherein the input of the model is the embedding representation of the user social behavior feature vector and the coupon label attributes, and the output is the matching result. 6.The social network behavior based electronic coupon recommendation acquisition method of claim 1, wherein, In step S6, the feedback behavior data comprises click, collection, sharing and redemption behaviors on the recommended electronic coupons.
7. An electronic coupon recommendation acquisition system based on social network behavior, characterized by, A device for implementing the method of any one of claims 1-6, comprising: a data collection module for safely collecting multi-dimensional social behavior data of a target user through a social network platform; a data processing and analysis module for cleaning, structuring and feature extraction of the social behavior data, and calculating a social influence index and a dynamic interest preference graph of the user; a coupon management module for maintaining a candidate electronic coupon set and its label attributes; The intelligent recommendation module has a built-in recommendation model, which is used to fuse user social behavior analysis results, coupon attributes and collaborative filtering signals to generate a personalized electronic coupon recommendation list; The push and feedback module is used to push the recommendation list to target users and collect user feedback behavior data; The model optimization module is used to continuously train and update the recommendation model according to the feedback behavior data. 8.The social network behavior based electronic coupon recommendation acquisition system of claim 7, wherein, The data collection module includes a privacy management unit, which is used to verify the user authorization state before collecting data and to anonymize and desensitize the collected personal information. 9.The social network behavior based electronic coupon recommendation acquisition system of claim 7, wherein, The data processing and analysis module includes: A feature extraction unit is used to convert unstructured social behavior data into a feature vector; An influence calculation unit quantifies user social influence based on a graph calculation algorithm; An interest modeling unit constructs a dynamic interest preference graph based on natural language processing and time series analysis. 10.A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the method according to any one of claims 1-6 when executing the computer program.