Dynamic NPRank algorithm based on authentication characteristics of multiple social platforms and method for evaluating influence of social network users by using dynamic NPRank algorithm

Abstract

The invention discloses a dynamic NPRank algorithm based on authentication characteristics of multiple social platforms and a method for evaluating the influence of social network users by using the dynamic NPRank algorithm. In order to solve the problems of single influence evaluation dimension and insufficient algorithm precision in the prior art, the invention provides a dynamic NPRank algorithm based on multiple social platform authentication features, and on the basis of a PeopleRank algorithm, a damping factor is adjusted according to whether a user passes platform authentication or not so as to distinguish influence propagation weights of different types of users; the invention further provides a method for evaluating the influence of the social network users according to an NPRank algorithm. The method comprises the following steps: collecting public data of a target user from a social network platform; establishing an attention relation probability model; constructing an adjacent matrix of the user relationship; the influence is calculated in parallel; and fusing calculation results to obtain a comprehensive influence score. The method is suitable for social data analysis fields such as brand marketing, public opinion monitoring, content propagation research and key opinion screening.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of social network data analysis and user influence evaluation, and particularly relates to a user influence evaluation method combining multi-social platform data and multi-algorithm fusion calculation and a dynamic NPRank algorithm thereof. BACKGROUND

[0002] With the popularization and commercial development of social network platforms, user influence evaluation has gradually become an important tool for brand promotion, public opinion analysis and key opinion leader screening. Current mainstream evaluation methods mostly rely on data statistics of a single platform, such as basic indicators such as the number of fans, interaction volume or the number of works. Although such methods have low implementation costs, they are difficult to reflect the real propagation ability of users in the social relationship network, especially in the presence of "fake fans" or "zombie fans", the evaluation results often deviate from the actual influence.

[0003] Existing algorithm models are mostly represented by fixed damping factor algorithms such as PageRank, which have obvious limitations in universality and scalability. Traditional algorithms fail to fully consider the differences in user attributes, such as the natural difference in propagation weight between authenticated users and ordinary users, resulting in insufficient influence evaluation accuracy. In addition, a single algorithm model is difficult to adapt to the data characteristics of different platforms and cannot cope with the complex relationship structure after the fusion of multi-source social platform data.

[0004] In the data application layer, existing evaluation systems generally lack unified cross-platform data collection capabilities and visualization display functions. The evaluation results are mostly presented in the form of static tables, lacking multi-dimensional graphical expression, making it difficult to provide intuitive comparison and insight for analysts. In addition, the evaluation process lacks real-time log recording and automatic export mechanism, making the algorithm execution process untraceable, which is not conducive to result verification and subsequent analysis reuse.

[0005] Therefore, there is an urgent need for a social network user influence evaluation method that can fuse multi-platform data, introduce a dynamic weight adjustment mechanism and have visualization display and process closed loop characteristics to solve the deficiencies of existing technologies in precision, adaptability and operability. SUMMARY

[0006] In order to solve the problems of single influence evaluation dimension, insufficient algorithm accuracy and lack of visualization and process closed loop in the prior art, the present application proposes the following solutions: A dynamic NPRank algorithm based on multi-social platform authentication features, on the basis of the PeopleRank algorithm, adjusts the damping factor according to whether the user is authenticated by the platform to distinguish the influence propagation weight of different types of users, specifically including: For users who have not passed platform authentication, calculate their influence score according to the following formula: , For users who have passed platform authentication, the influence score is calculated as follows: , Wherein, is the damping factor of users who have not passed platform authentication, is the damping factor of users who have passed platform authentication, is the user to be evaluated, is the user 's fan, is the node set linked to the user , is the total number of users followed.

[0007] A method for evaluating the influence of a social network user, comprising: S1, collecting public data of a target user from a social network platform to obtain a user data table containing user basic information; S2, establishing a follow relationship probability model according to the user data table, and determining the connection relationship strength between users according to the probability model; S3, constructing an adjacency matrix of user relationship according to the connection relationship strength, for representing the follow association between users; S4, taking the adjacency matrix as input, respectively performing influence calculation based on PeopleRank algorithm and influence calculation based on the dynamic NPRank algorithm of the application; S5, fusing the calculation results of the two algorithms in S4 and writing them into the user data table to obtain a comprehensive influence score.

[0008] Further, the user basic information includes user ID, nickname, number of follows, number of fans, number of works, and whether it has passed platform authentication.

[0009] Further, in S2, the follow relationship probability model determines the connection probability according to the proportion of the number of fans of each user, and the more the number of fans, the higher the probability of being followed.

[0010] Further, in S3, the matrix elements of the adjacency matrix are used to represent the follow relationship between users, and the elements in the matrix take value 1 when there is a follow relationship and take value 0 when there is no follow relationship.

[0011] Further, in S4, the iteration calculation of the PeopleRank algorithm is performed with a fixed damping factor as a parameter, and the dynamic NPRank algorithm selects different damping factors for calculation according to whether the user has passed platform authentication.

[0012] Further, in S5, the calculation results of the two algorithms are fused by using a weighted average method to obtain a fusion result.

[0013] Further, the comprehensive influence score is displayed by visualization, and the visualization result includes a ranking comparison chart, a user multi-dimensional radar chart and a user relationship network chart, so as to intuitively display the influence difference between users.

[0014] Further, the comprehensive influence score is supported to be exported as a spreadsheet file, and log records are generated, the log records including data collection time, algorithm parameters and abnormal information.

[0015] Further, during the running of the method, the task progress, the current processing user and the iteration number are displayed in real time, and prompt information is output when input errors or access abnormalities are detected.

[0016] Compared with the prior art, the present application has the following beneficial effects: The dynamic NPRank algorithm proposed in the present application dynamically adjusts the damping factor according to whether the user is authenticated by the platform on the basis of the traditional PeopleRank algorithm, so as to distinguish the propagation weights of different types of users. The improved algorithm can more accurately reflect the difference between the authority of authenticated users and the activity of ordinary users, and improve the distinguishability and reliability of the influence score.

[0017] The social network user influence evaluation method based on multi-algorithm fusion provided by the present application realizes unified acquisition and standardized processing of user information of mainstream social platforms such as Douyin, Weibo and Zhihu by establishing a cross-platform data acquisition mechanism. This method effectively avoids the deviation caused by single platform data, so that the evaluation result can more comprehensively reflect the real influence of users in different social ecologies.

[0018] The present application quantifies the fan relationship between users into a calculable network structure by introducing a follow relationship probability model and an adjacency matrix construction technology. In the algorithm calculation, the matrix is used as the basis, which significantly improves the expression ability of the model to the real social interaction relationship, and reduces the interference of false fans and surface interaction on the result.

[0019] The present application realizes multi-algorithm collaborative optimization through the fusion calculation of PeopleRank and dynamic NPRank algorithms. Avoiding the error caused by parameter solidification or data skew of a single algorithm, the influence evaluation result is more stable and fair among different user groups.

[0020] The application presents the influence distribution and social structure characteristics of users intuitively through multi-dimensional visualization modes such as ranking comparison charts, radar charts and relationship network charts. The scheme improves the interpretability of the results, so that data analysts can quickly identify key nodes and potential high-influence users.

[0021] In addition, the application has perfect export and logging functions, supports outputting the evaluation results in a spreadsheet format, and automatically records running parameters and abnormal information, which facilitates subsequent data tracing and system optimization. This feature forms a closed loop for the entire evaluation process, improving the reliability and engineering practicability of the system.

[0022] The application has the characteristics of strong multi-platform adaptability, high algorithm evaluation accuracy and intuitive result display, and can realize comprehensive analysis and objective evaluation of the influence of social network users, and is suitable for social data analysis fields such as brand marketing, public opinion monitoring, content dissemination research and key opinion screening. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is a flow chart of a method for evaluating the influence of social network users according to an embodiment; Figure 2 is a data crawling flow chart of a method for evaluating the influence of social network users according to an embodiment; Figure 3 is an influence calculation flow chart of a method for evaluating the influence of social network users according to an embodiment; Figure 4 is a system overall architecture diagram of a method for evaluating the influence of social network users according to an embodiment. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the application will be described below in detail with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0025] Embodiment one A dynamic NPRank algorithm based on multi-social platform authentication features, on the basis of the PeopleRank algorithm, adjusts the damping factor according to whether the user has passed platform authentication, to distinguish the influence propagation weight of different types of users, specifically including: For users who have not passed platform authentication, calculate their influence score according to the following formula: , For users who have passed platform authentication, calculate their influence score according to the following formula: , in, It is the damping factor of users who have not passed platform authentication. It is the damping factor of users who have been certified by the platform. Users awaiting evaluation User fans, It is a chain to users The set of nodes, yes Total number of users following.

[0026] This implementation introduces a dynamic NPRank algorithm based on authentication features across multiple social platforms on top of the PeopleRank algorithm. By adjusting the damping factor according to whether a user has been authenticated by the platform, the algorithm can distinguish the propagation weight of different types of users, thereby more accurately reflecting the difference in influence between authenticated users and ordinary users in social networks and improving the accuracy and credibility of the evaluation results.

[0027] like Figure 1 As shown, a method for assessing the influence of social network users includes: S1. Collect publicly available data of target users from social networking platforms to obtain a user data table containing basic user information; S2. Based on the user data table, establish a probability model of attention relationship, and use the probability model to determine the strength of the connection relationship between users; S3. Construct an adjacency matrix of user relationships based on the strength of the connection relationship, which is used to represent the attention associations between users; S4. Using the adjacency matrix as input, perform influence calculation based on the PeopleRank algorithm and influence calculation based on the dynamic NPRank algorithm of the present invention, respectively. S5. The calculation results of the two algorithms described in S4 are fused and written into the user data table to obtain the comprehensive influence score.

[0028] This implementation method automates the entire process from data collection, relationship modeling, algorithm calculation to result output by constructing a complete social network user influence assessment process. It can uniformly assess user influence in a multi-platform environment, effectively improving data processing efficiency and assessment accuracy.

[0029] Furthermore, the basic user information includes user ID, nickname, number of followers, number of fans, number of works, and whether the user has been certified by the platform.

[0030] This implementation method ensures the comprehensiveness and consistency of algorithm input by including core information such as user ID, nickname, number of followers, number of fans, number of works, and authentication status in the user data table, thus providing high-quality data support for subsequent relationship modeling and algorithm calculation.

[0031] Furthermore, in S2, the attention relationship probability model determines the connection probability based on the proportion of each user's followers, with users having more followers having a higher probability of being followed.

[0032] This implementation method establishes a probability model by using the proportion of fans during the follower relationship modeling stage, so that users with more fans have a higher probability of being followed. This statistically simulates the structure of real social networks and improves the fit between the generated relationship network and actual social behavior.

[0033] Furthermore, in S3, the matrix elements of the adjacency matrix are used to represent the attention relationship between users. When an element in the matrix takes a value of 1, it indicates that there is an attention relationship, and when it takes a value of 0, it indicates that there is no attention relationship.

[0034] This implementation constructs an adjacency matrix to represent the attention relationships between users, transforming complex social interactions into a computable matrix form. This enables subsequent algorithms to perform efficient iterative calculations at the mathematical level, thereby improving system processing efficiency and model stability.

[0035] Furthermore, in S4, the iterative calculation of the PeopleRank algorithm is performed with a fixed damping factor as a parameter, while the dynamic NPRank algorithm selects different damping factors for calculation based on whether the user has passed platform authentication.

[0036] This implementation distinguishes between the PeopleRank algorithm with fixed damping and the NPRank algorithm with dynamic damping during the algorithm calculation stage. This allows the model to retain the convergence stability of the traditional algorithm while introducing the flexibility of user attribute differences, thereby achieving a more reasonable distribution of influence among different types of users.

[0037] Furthermore, in S5, when the calculation results of the two algorithms are fused, a weighted average method is used to obtain the fusion result.

[0038] This implementation method effectively reduces the impact of single algorithm bias by using a weighted average algorithm to integrate the results during the fusion stage, making the comprehensive influence score more representative and robust, and improving the overall consistency of the evaluation results.

[0039] Furthermore, the comprehensive influence score is displayed in a visual manner, and the visualization results include a ranking comparison chart, a user multi-dimensional radar chart, and a user relationship network diagram, which are used to intuitively show the differences in influence among users.

[0040] This implementation method uses multi-dimensional visualization methods such as ranking comparison charts, radar charts, and relationship network diagrams to make user influence results more intuitive and visible, helping analysts quickly identify key nodes and user structure characteristics, and improving the interpretability and decision-making value of the results.

[0041] Furthermore, the comprehensive influence score can be exported as a spreadsheet file and a log record can be generated, which includes data collection time, algorithm parameters, and anomaly information.

[0042] This implementation method ensures that the evaluation data can be stored and traced for a long time by adding result export and log recording functions to the system, which facilitates subsequent analysis, report generation and algorithm optimization, and improves the system's practicality and maintainability.

[0043] Furthermore, during the operation of the method, the task progress, the current user being processed, and the number of iterations are displayed in real time, and a prompt message is output when an input error or access abnormality is detected.

[0044] This implementation method improves the stability of the evaluation process and enhances the user experience by displaying task progress, current calculation status, and error prompts in real time during operation.

[0045] Implementation Method 2 This embodiment combines the technical solutions described in the foregoing embodiments, and, in conjunction with actual application scenarios and the usage process of computer program products that implement the method of this invention, further verifies and explains the technical effects of this invention through specific examples.

[0046] S1: Multi-platform social data crawling This invention supports the collection of publicly available data from platforms such as Douyin, Weibo, and Zhihu, and allows for the configuration of specific parameters based on the platform type. like Figure 2 As shown, this implementation method takes Douyin as an example, and obtains Douyin's VerifyFP through browser F12->Application->Cookies; Enter a list of user IDs / nicknames, separated by commas. Set the crawling interval, which is 2 seconds by default, to avoid platform API limitations. The core data fields to be crawled are: user ID, nickname, number of followers, number of fans, whether the user is verified (1=verified, 0=normal), and number of works. These data are stored as structured data, and this implementation uses the DataFrame format.

[0047] S2: Construct an adjacency matrix for user relationships A probability model of follower relationship is constructed based on the crawled "number of followers": the more followers a user has, the higher the probability that other users will follow them. In this implementation, the number of followers is taken as the minimum value of 1 to avoid zero value.

[0048] Each user is limited to following 2 to 5 other users, and self-following is prohibited, excluding the user's own user ID; Generate an N×N adjacency matrix (N is the number of users). The matrix element "1" indicates that "user i follows user j" and "0" indicates that there is no follow relationship. The matrix is ​​stored on an available GPU to accelerate subsequent calculations.

[0049] S3: The Impact of Multi-Algorithm Fusion Computation like Figure 3 As shown, PeopleRank is calculated as follows: the initial influence value (PR) is 1 / N. In this preferred embodiment, the damping factor is d=0.85, the maximum number of iterations is 100, and the convergence threshold is... The calculation is performed iteratively using the following formula until convergence (the difference in PR between adjacent iterations < the threshold).

[0050] The PeopleRank algorithm is based on the PageRank model. Its core is to calculate influence through the "following-follower" link relationship between users, as shown in the following formula:

[0051] Symbol definition: :user PeopleRank score (i.e., influence score); All user nodes in a social network; : Link to users The set of nodes (i.e.) (fan base) :user The number of fans (i.e.) (size of the set) Damping factor (default 0.85, representing the probability that a user will continue to "follow" the link rather than randomly following new users).

[0052] NPRank calculation: Damping factors are set according to user authentication status (d1=0.85 for ordinary users, d2=0.95 for authenticated users); it is calculated by formula PR=(1-d_values)+Σ[(PR_old / out-degree)×adjacency matrix transpose×d_values] (d_values ​​is the exclusive damping factor for each user). The core improvement of the dynamic NPRank algorithm in this implementation is that it adjusts the damping factor based on whether the user has passed "Douyin personal verification" to weaken the influence of "zombie followers". The formula is divided into two scenarios: 1. Scenario 1: Regular registered user (unverified) )

[0053] 2. Scenario 2: Personally verified user (already verified, )

[0054] Symbol definition: :user NPRank value (optimized influence score); User authentication status identifier ( Regular users Authenticated users); Damping factor for ordinary users (set to 0.85 in the document); Damping factor for certified users (set to 0.95 in the document, as certified users have higher credibility and therefore a higher weight in the "follow" link). : Link to users The set of nodes (similar to the "fan base" in PeopleRank, i.e.) yes (fans) :user The number of outgoing chains (i.e.) The total number of users followed is used for equal distribution. Influence weight).

[0055] Final Influence Score: In this implementation, the average of the PeopleRank and NPRank results is taken, that is, the weight of the results of the two algorithms is 0.5, and then updated to the original data.

[0056] S4: Multi-dimensional Results Visualization Generate three types of visualization charts: Ranking Comparison Bar Chart: Sorted by PeopleRank, comparing the ranking of the same user in two algorithms (the smaller the ranking value, the greater the influence; Y-axis inverted). User Multidimensional Radar Chart: Select the top 5 users and display normalized indicators (values ​​0-1) for "number of followers, number of followings, number of works, whether verified, and influence score". User relationship network diagram: Node size corresponds to influence score, edges represent attention relationships, and the network structure is displayed using a spring layout.

[0057] S5: Results Export and Log Recording Supports exporting results to Excel (multi-sheet: original data, PeopleRank results, NPRank results) or CSV format; Generate a system log (influence_system.log) to record crawling time, calculation parameters, and exception information, and display it in real time on the GUI log tab.

[0058] S6: Status Control and Exception Handling The progress bar (0-100%) is updated in real time during the crawling process, displaying the current crawling user; The number of iterations is recorded during the calculation (e.g., "PeopleRank converged on the 15th iteration"), and a pop-up message is displayed when an error occurs (e.g., "No user ID entered" or "VerifyFP error").

[0059] The overall system architecture of this embodiment is as follows: Figure 4 As shown, the overall system architecture is divided into three layers: data layer (multi-platform data crawling, adjacency matrix), algorithm matrix, algorithm layer (PeopleRank and NPRank fusion), and presentation layer (GUI tabs: data crawling, data analysis, result display, and system logs).

[0060] The above detailed description of the technical solution provided by the present invention is intended to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above detailed embodiments are not intended to limit the scope of protection of the present invention. Any reasonable modifications and improvements to the present invention, recombination of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0061] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims disclosed in the present invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle scope of the present invention should be considered to fall within the protection scope of the present invention.

[0062] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A dynamic NPRank algorithm based on authentication features from multiple social platforms, characterized in that, Based on the PeopleRank algorithm, a damping factor is adjusted according to whether a user has passed platform verification to differentiate the influence propagation weight of different types of users, specifically including: For users who have not passed platform verification, their influence score is calculated using the following formula: ， For users who have been verified by the platform, their influence score is calculated using the following formula: ， in, It is the damping factor of users who have not passed platform authentication. It is the damping factor of users who have been certified by the platform. Users awaiting evaluation User fans, It is a chain to the user The set of nodes, yes Total number of users following.

2. A method for assessing the influence of social network users, characterized in that, include: S1. Collect publicly available data of target users from social networking platforms to obtain a user data table containing basic user information; S2. Based on the user data table, establish a probability model of attention relationship, and use the probability model to determine the strength of the connection relationship between users; S3. Construct an adjacency matrix of user relationships based on the strength of the connection relationship, which is used to represent the attention associations between users; S4. Using the adjacency matrix as input, perform influence calculation based on the PeopleRank algorithm and influence calculation based on the dynamic NPRank algorithm of claim 1, respectively. S5. The calculation results of the two algorithms described in S4 are fused and written into the user data table to obtain the comprehensive influence score.

3. The method according to claim 2, characterized in that, The basic user information includes user ID, nickname, number of followers, number of fans, number of works, and whether the user is verified by the platform.

4. The method according to claim 2, characterized in that, In S2, the probability model of the following relationship determines the connection probability based on the proportion of each user's number of followers. The more followers a user has, the higher the probability of being followed.

5. The method according to claim 2, characterized in that, In S3, the matrix elements of the adjacency matrix are used to represent the attention relationship between users. When an element in the matrix takes a value of 1, it indicates that there is an attention relationship, and when it takes a value of 0, it indicates that there is no attention relationship.

6. The method according to claim 2, characterized in that, In S4, the iterative calculation of the PeopleRank algorithm is performed with a fixed damping factor as a parameter, while the dynamic NPRank algorithm selects different damping factors for calculation based on whether the user has passed platform authentication.

7. The method according to claim 2, characterized in that, In S5, when the calculation results of the two algorithms are fused, a weighted average method is used to obtain the fusion result.

8. The method according to claim 2, characterized in that, The comprehensive influence score is displayed in a visual manner. The visualization results include a ranking comparison chart, a multi-dimensional radar chart of users, and a user relationship network chart, which are used to intuitively show the differences in influence among users.

9. The method according to claim 2, characterized in that, The comprehensive influence score can be exported as a spreadsheet file and a log record can be generated, which includes the data collection time, algorithm parameters and anomaly information.

10. The method according to claim 2, characterized in that, During the operation of the method, the task progress, the current user being processed, and the number of iterations are displayed in real time. When an input error or access abnormality is detected, a prompt message is output.

Images