User portrait enhancement method based on dynamic tag

Abstract

The invention relates to the technical field of user data processing, in particular to a user portrait enhancement method based on a dynamic tag, which is applied to a system method for designing data acquisition, data processing, data analysis and data classification of a user information and user portrait data use platform. By analyzing the user behavior in real time and dynamically adjusting the label in the user portrait, the more accurate and comprehensive user portrait can be constructed, and the real-time interest and long-term preference of the user can be captured. Based on the dynamically updated user portraits, the system can provide highly personalized recommendation services, commodity recommendation, content recommendation or service recommendation can be closer to the real demands of the users, and the satisfaction degree and participation degree of the users are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of user data processing, and specifically to a user portrait enhancement method based on dynamic tags, which is applied to a systematic method for designing user information and user portrait data using a platform for data collection, data processing, data analysis, and data classification. Background Art

[0002] With the rapid development of the Internet and big data technology, personalized recommendation systems have been widely used in various fields. As the core component of personalized recommendation systems, the accuracy and completeness of user portraits are crucial to system performance. However, traditional user portrait construction methods often cannot fully reflect users' real-time interests and behaviors, and a more flexible and dynamic method is needed to enhance user portraits.

[0003] However, traditional user portrait construction methods often fail to accurately reflect users' real-time interests and behaviors. They only display basic information about users and cannot meet some special needs. Therefore, a more flexible and dynamic method is needed to enhance user portraits. Summary of the invention

[0004] Traditional methods of building user profiles usually rely on static features, such as the user's age, gender, and past purchase records. Although this method can make recommendations based on the user's basic information and past behavior, it may not be able to capture the user's latest interests and needs in a fast-changing market environment, resulting in low accuracy of recommendations. For example, if a user has recently started to pay attention to healthy foods, but this change is not reflected in his or her historical purchase records, then the traditional recommendation system may still recommend non-related products to him or her based on the old preferences.

[0005] User portrait enhancement method based on dynamic tags: In order to solve the above problems, we proposed a user portrait enhancement method based on dynamic tags. The core of this method is to analyze the user's activity data in real time, including browsing behavior, search keywords, click feedback, etc., and extract dynamic tags that reflect the user's current interests, so as to update the user portrait and make it closer to the user's immediate needs.

[0006] Take a user as an example. Suppose that the user has recently frequently browsed the sports equipment page and searched for "outdoor running shoes". Traditional methods may ignore these emerging points of interest due to the lack of real-time update mechanism, but after the dynamic tag enhancement, the system can quickly capture these signals and tag users with dynamic tags such as "sports" and "running shoes". Subsequently, the recommendation algorithm will give priority to recommending sports and running-related products to the user, rather than just based on past shopping records. This not only improves the relevance of the recommendation, but also increases user satisfaction and purchase likelihood.

[0007] We conducted a series of experiments on real datasets to compare the effects of user profile enhancement methods based on dynamic tags and traditional methods. The results show that the dynamic tag method can significantly improve the accuracy of recommendations.

[0008] A user portrait enhancement method based on dynamic tags includes the following steps: 1. The data collection module collects data without affecting the user experience by embedding the system code. First, it is necessary to collect user behavior data, interest preference information, etc. from various data sources. These data may include user click records, purchase records, search records, etc., which are used to analyze user behavior characteristics and interest points.

[0009] 2. Store the collected data into the corresponding table of the database through asynchronous methods.

[0010] When processing large amounts of real-time data, using asynchronous methods to store data in the database improves the system's response speed and processing capabilities. Here we take the Java language as an example, combined with the Spring Boot framework and the Reactor library; Use Spring Data JPA to define the data model; write data asynchronously Next, use the Mono or Flux type in Reactor to handle asynchronous operations and insert multiple data asynchronously at one time; 3. Calculate the importance of the collected user data and historical behaviors through a dynamic tag importance generation algorithm; n(i,j) is the number of times keyword j appears in document i, |D| is the total number of documents in the corpus, and |j:t(i)∈d(j)| is the total number of documents in which word i appears; It is necessary to build an inverted index to record which documents each keyword appears in and how many times it appears; calculate the total number of documents and keyword frequency; traverse the entire corpus and count the total number of documents (|D|) and the number of keywords (n(i,j)); For each keyword, use the above formula to calculate its The value is then stored in the corresponding user behavior data; 4. Generate corresponding dynamic labels based on the calculated label importance data; A threshold is set according to business needs. Only tags with importance exceeding the threshold are retained. Tags with high importance are selected from the calculation results to form a dynamic tag set. Duplicate tags are removed to ensure that the naming and format of all tags are consistent. The generated dynamic tags are stored in a cache or database. 5. Match and update the dynamic tags generated by users with the user portrait to achieve dynamic enhancement of the user portrait. The user portrait update algorithm updates the user portrait’s points of interest, behavioral characteristics, and other content based on the user’s dynamic tag information; Get the latest generated dynamic tags from the real-time data stream or cache or database; Read user profile data from the database or persistent storage; Ensure that the user profile is stored in a structured format for easy modification and updating; Compare the existing tags in the user profile with the newly generated dynamic tags to find similar or related tags; Decide whether to update the points of interest and behavioral characteristics in the user profile based on the importance and relevance of the dynamic tags; Integrate the successfully matched dynamic tags into the user profile, including adding new tags, updating the weight of existing tags, or removing tags that are no longer active; Write the updated user profile back to the database or persistent storage. Publish user profile update events through the event bus or message queue.

[0011] Beneficial Effects of the Invention 1. Accurate user portrait construction: By analyzing user behavior in real time and dynamically adjusting the labels in the user portrait, a more accurate and comprehensive user portrait can be constructed to capture the user's real-time interests and long-term preferences.

[0012] 2. Personalized recommendation service: Based on dynamically updated user portraits, the system can provide highly personalized recommendation services. Whether it is product recommendations, content recommendations or service recommendations, they can be closer to the real needs of users and improve user satisfaction and engagement.

[0013] 3. Marketing strategy optimization: Companies can use the deep insights provided by dynamic user portraits to develop more precise marketing strategies, such as personalized advertising and targeted promotions, to improve marketing efficiency and conversion rates.

[0014] 4. Product and service innovation: By analyzing the changing trends of user portraits, companies can discover new market opportunities, provide a basis for product innovation and service improvement, and maintain competitive advantage.

[0015] 5. Improved user experience: The dynamic tagging mechanism can ensure that the recommended content is highly matched with the user's current status, providing a smooth, consistent and personalized experience whether in shopping, reading or entertainment scenarios.

[0016] 6. Data-driven decision support: Dynamic user portraits not only serve personalized recommendations, but are also an important source of data for companies to conduct strategic planning, market analysis and other decision-making processes, helping companies make more informed decisions.

[0017] This patented solution effectively improves the accuracy and effectiveness of personalized recommendations by analyzing and updating user interests in real time, providing users with better services while also bringing higher commercial value and market competitiveness to enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The figure is a flow chart of a user portrait enhancement method based on dynamic tags. DETAILED DESCRIPTION

[0019] The present invention provides a user portrait enhancement method based on dynamic tags, which solves the above problems.

[0020] In order to solve the above technical problems, the present invention is achieved through the following technical solutions: The present invention is a user portrait enhancement method based on dynamic tags, comprising a data acquisition module, a dynamic tag generation module and a user portrait update module.

[0021] 1. Data acquisition module Function: Collect user data from various sources, including but not limited to user's basic information, historical behavior data, social network activities, etc.

[0022] Technical requirements: Efficient data capture and cleaning capabilities to ensure data quality.

[0023] Example: Use crawler technology to obtain user comments from websites, and use API interfaces to obtain user interaction data from social media platforms.

[0024] 2. Dynamic label generation module Function: Dynamically generate tags describing users based on real-time or near real-time user behavior.

[0025] Technical requirements: Real-time data analysis and processing capabilities, able to quickly respond to changes in user behavior.

[0026] Example: When a user browses multiple articles about "fitness" in succession, the system automatically adds the "fitness" tag.

[0027] 3. User portrait update module Function: Based on dynamic tags, user portraits can be updated in real time or regularly to ensure their accuracy and timeliness.

[0028] Technical requirements: Efficient database operations and data processing algorithms to support rapid updates of large-scale user data.

[0029] Example: Run the script every night to update the interest field in the user portrait based on the dynamic tags of the day.

[0030] Specific process of implementation: 1. The data collection module collects data without affecting the user experience by embedding the system code. First, it is necessary to collect user behavior data, interest preference information, etc. from various data sources. These data may include user click records, purchase records, search records, etc., which are used to analyze user behavior characteristics and interest points.

[0031] 1. Design and deployment of tracking code Function: Insert lightweight monitoring code (i.e. "embedding") at key user interaction points, such as page loading, button clicks, form submissions, and other events, to capture user behavior data.

[0032] Technical requirements: Use non-blocking asynchronous calling methods to avoid affecting page loading speed or response time. Ensure that the embedding code takes up less resources and does not cause page lag. Support cross-platform and cross-device compatibility, including but not limited to Web and mobile App.

[0033] Example: When a user clicks the "Add to Cart" button, a tracking event is triggered through JavaScript to record information such as the product ID and click timestamp.

[0034] 2. Data structuring and storage Function: Convert the collected raw data into a structured format to facilitate subsequent analysis and processing.

[0035] Technical requirements: For structured data (such as database query results), directly use CSV, JSON and other formats for storage. For unstructured data (such as user comments, forum posts), use TXT or JSON format for storage.

[0036] Example: Use Java's MySql library to read user purchase records from the database and convert them into DataFrame format for further statistical analysis.

[0037] 3. Real-time data transmission and processing Function: Ensure that data can be transmitted to the data warehouse or analysis server in real time or near real time so that user profiles can be updated in a timely manner.

[0038] Technical requirements: Use message queues (such as Kafka, RabbitMQ) to process data streams to ensure the stability and high throughput of data transmission. Implement data cleaning and preprocessing logic to eliminate invalid or abnormal data.

[0039] Example: Every time a user completes a search operation, the corresponding tracking data is transmitted to the data processing pipeline in real time through Kafka. After filtering and formatting, it is stored in Hadoop HDFS or similar big data storage systems. In the early stages of the project, the MySql database can be used directly for storage.

[0040] 2. Further store the collected data into the corresponding table of the database through asynchronous methods.

[0041] Technical implementation process: When processing large amounts of real-time data, using asynchronous methods to store data in the database can significantly improve the system's response speed and processing capabilities. Here we take the Java language as an example, combined with the Spring Boot framework and the Reactor library.

[0042] Step 1: Define the data model First, define the data model using Spring Data JPA.

[0043] Step 2: Write data asynchronously Next, use the Mono or Flux types in Reactor to handle asynchronous operations.

[0044] Step 3: Batch asynchronous insertion To improve efficiency, multiple data can be inserted asynchronously at one time.

[0045] 3. Further calculate the importance of the collected user data and historical behaviors through a dynamic tag importance generation algorithm.

[0046] n(i,j): The number of times keyword j appears in document i.

[0047] |D|: total number of documents in the corpus |j:t(i) belongs to d(j)|The total number of documents in which word i appears Technical implementation process: Step 1: Build an inverted index First, you need to build an inverted index to record in which documents each keyword appears and the number of times it appears.

[0048] Step 2: Calculate the total number of documents and keyword frequency Traverse the entire corpus and count the total number of documents ( |D| ) and the number of keywords (n(i,j)).

[0049] Step 3: Calculation value For each keyword, use the above formula to calculate its The value is then stored in the corresponding user behavior data.

[0050] Fourth, corresponding dynamic labels can be further generated according to the calculated label importance data.

[0051] Technical implementation process: Step 1: Dynamic label generation Threshold setting: Set a threshold according to business needs, and only tags whose importance exceeds the threshold will be retained.

[0052] Label generation: Filter out labels with high importance from the calculation results to form a dynamic label set.

[0053] Step 2: Label Standardization Deduplication: Remove duplicate tags to keep the tag set unique.

[0054] Unified format: Ensure that all tags are named and formatted consistently to facilitate subsequent processing.

[0055] Step 3: Storing dynamic tags Storage location: Store the generated dynamic tags in a cache or database for quick access.

[0056] Update mechanism: Design a reasonable update strategy, such as excluding tags with importance lower than 0.06, and retaining only the top ten by importance after exclusion, to ensure that dynamic tags can reflect the latest user behavior patterns in a timely manner.

[0057] 5. Further match and update the dynamic tags generated by users with the user portrait to achieve dynamic enhancement of the user portrait. The user portrait update algorithm can update the points of interest, behavioral characteristics and other content in the user portrait based on the user's dynamic tag information.

[0058] Technical implementation process: Step 1: Get dynamic tags Data source: Get the latest generated dynamic labels from real-time data streams or caches or databases.

[0059] Data format: Make sure dynamic tags are stored in an easy to process format, such as a dictionary or structured object.

[0060] Step 2: Load the user profile Data source: Read user profile data from a database or persistent storage.

[0061] Data format: Ensure that user profiles are stored in a structured format to facilitate modification and updating.

[0062] Step 3: Matching user profiles with dynamic tags Matching logic: Compare the existing tags in the user portrait with the newly generated dynamic tags to find similar or related tags.

[0063] Update rules: Based on the importance and relevance of dynamic tags, decide whether to update the points of interest and behavioral characteristics in the user portrait.

[0064] Step 4: Update the user profile Update operation: Incorporate successfully matched dynamic tags into the user profile, which may include adding new tags, updating the weight of existing tags, or removing tags that are no longer active.

[0065] Consistency check: Ensure that the update operation does not destroy the consistency and integrity of the user profile.

[0066] Step 5: Store the updated user profile Data storage: Write the updated user profile back to the database or persistent storage.

[0067] Version Control: Consider implementing a version control mechanism to track the change history of your user personas.

[0068] Step 6: Notify downstream systems Event publishing: Publish user profile update events through the event bus or message queue to notify downstream systems such as the recommendation system and advertising system.

[0069] Synchronous updates: Ensure that all systems that rely on user portraits can obtain updated information in a timely manner.

[0070] Through real-time updating and adjustment of dynamic tags, the user portrait can be accurately described, thereby improving the accuracy and effectiveness of personalized recommendations and making the user portrait more real and comprehensive. The dynamic tag mechanism can quickly capture and reflect new behaviors and preference changes of users. For example, if a user recently starts to pay attention to a certain type of product or service, the system can instantly adjust its user portrait to provide recommendations that are more in line with current interests. Compared with static portraits that are updated regularly, the dynamic mechanism greatly reduces the time difference from the occurrence of behavior to the update of the portrait, ensuring the timeliness and relevance of the recommended content. Based on more accurate user portraits, companies can provide more personalized product, service or content recommendations to meet the diverse and personalized needs of users and improve user satisfaction and loyalty. Based on dynamic user portraits, companies can formulate more accurate marketing strategies, such as targeted advertising, personalized promotional activities, etc., to improve the efficiency and return on investment of marketing resources.

[0071] 1. Data collection: Collect user behavior data, interest preference information, etc. from various data sources. These data may include user click records, purchase records, search records, etc., and store the collected data in the corresponding table of the database through asynchronous methods.

[0072] Randomly extract click records, purchase records, search records and other data collected from a user within a week; 2. Importance calculation: The importance of collected user data and historical behaviors is calculated through a dynamic tag importance generation algorithm.

[0073] The following are the top five keywords calculated based on the database collection content: - computer: = 0.093 * 1.386 = 0.129 - java: = 0.062 * 1.386 = 0.086 - cell phone: = 0.058 * 1.386 = 0.080 - watch: = 0.031 * 1.386 = 0.043 - brand: = 0.023 * 1.386 = 0.032 1. Generate dynamic tags Those with importance lower than 0.06 are not generated. Therefore, three dynamic tags are generated: computer, java, and mobile phone.

[0074] 2. Dynamic tags are matched and updated with user portraits The user portrait tag display page only displays the top ten tags in terms of importance. All newly generated tags and the user's original tags are ranked in order of importance, and the top ten are displayed. The above number of items and data are for illustration only and can be adjusted based on the specific collected data. This is only for illustration of matching updates.

[0075] In summary, the user portrait enhancement method based on dynamic tags effectively improves the accuracy and effect of personalized recommendations by real-time analysis and updating of user interests, providing users with better services. Different from the regular update of traditional static user portraits, the dynamic tag mechanism allows the system to dynamically adjust the tag weight and content according to the user's latest behavior, ensuring that the user portrait always reflects the user's current interests and needs. The user portrait enhancement method based on dynamic tags, with its real-time, high-precision and self-optimization characteristics, significantly improves the effectiveness of personalized recommendation systems and provides enterprises with deeper user insights and decision support. In the future, we will continue to optimize the algorithm and explore its application potential in different scenarios, in order to further promote the advancement of personalized recommendation technology. Through the above case analysis, we can clearly see that compared with traditional methods, the user portrait enhancement method based on dynamic tags has significant advantages in improving the timeliness and accuracy of personalized recommendations, and is a direction worthy of promotion and in-depth research.

Claims

1. A user portrait enhancement method based on dynamic tags, characterized in that The steps include:

1. The data collection module collects data without affecting the user experience by embedding the system code; First, it is necessary to collect user behavior data, interest preference information, etc. from various data sources; These data may include user click records, purchase records, search records, etc., which are used to analyze user behavior characteristics and interests; 2. Store the collected data into the corresponding table of the database through asynchronous methods; When processing large amounts of real-time data, using asynchronous methods to store data in the database improves the system's response speed and processing capabilities; Here we take Java language as an example, combined with Spring Boot framework and Reactor library; Use Spring Data JPA to define the data model; write data asynchronously Next, use the Mono or Flux type in Reactor to handle asynchronous operations and insert multiple data asynchronously at one time; 3. Calculate the importance of the collected user data and historical behaviors through a dynamic tag importance generation algorithm; is the number of times keyword j appears in document i, |D| is the total number of documents in the corpus, is the total number of documents in which word i appears; You need to build an inverted index to record which documents each keyword appears in and how many times it appears. Calculate the total number of documents and keyword frequency; traverse the entire corpus and count the total number of documents ( |D| ) and the number of keywords (n(i,j)); For each keyword, use the above formula to calculate its The value is then stored in the corresponding user behavior data; 4. Generate corresponding dynamic labels based on the calculated label importance data; A threshold is set according to business needs. Only tags with importance exceeding the threshold are retained. Tags with high importance are selected from the calculation results to form a dynamic tag set. Duplicate tags are removed to ensure that the naming and format of all tags are consistent. The generated dynamic tags are stored in a cache or database.

5. Match and update the dynamic tags generated by users with the user portraits to achieve dynamic enhancement of the user portraits; The user portrait update algorithm updates the user portrait’s interests, behavior characteristics, and other content based on the user’s dynamic tag information; Get the latest generated dynamic tags from real-time data streams, caches, or databases; Read user profile data from databases or persistent storage; Ensure that user profiles are stored in a structured format for easy modification and updating; compare existing tags in user profiles with newly generated dynamic tags to find similar or related tags; Decide whether to update the points of interest and behavior characteristics in the user portrait based on the importance and relevance of the dynamic tags; Integrate successfully matched dynamic tags into the user profile, including adding new tags, updating the weights of existing tags, or removing tags that are no longer active; write the updated user profile back to the database or persistent storage; and publish user profile update events through the event bus or message queue.

Images