A microblog group identification method based on community discovery

Abstract

This invention discloses a microblog group identification method based on community discovery, comprising the following steps: S1, data collection and cleaning; S2, feature extraction and representation; S3, establishing a classification model; S4, community tagging and influence analysis. In this invention, an optimized Dynamic Topic Model (DTM) is used to mine specific groups within the microblog community. Microblog posts from the past year are selected as the research object, and the similarity of topics in posts from different authors is used as the weight of links between authors, mapping the microblog network into a directed weighted network. Community discovery is performed using the Label Propagation Algorithm (LPA), identifying the inherent community structure within the social relationship network. This invention conducts in-depth analysis of user relationships within the microblog network, and based on identification methods for user-generated content characteristics, user association characteristics, and environmental characteristics, it mines potential topics to identify users with similar interests and active user groups in specific fields.

Description

Technical Field

[0001] This invention relates to the field of social network data analysis technology, and in particular to a method for identifying microblog groups based on community discovery. Background Technology

[0002] Currently, Sina Weibo, as a representative of online social media, has become the most mainstream online social platform in my country, boasting a vast user base and profound influence on public opinion. As of September 2021, Weibo had 573 million monthly active users and an average of 248 million daily active users. The "2020 Weibo User Development Report" released by Weibo shows that the Weibo user base continues to show a trend of becoming younger, with those born in the 1990s and 2000s accounting for nearly 80%.

[0003] Communities in microblogging networks are subgroups of users with the same or similar interests who can share and exchange information. Due to the characteristics of microblogging networks, such as large user scale, uneven degree distribution of nodes, and weak social networks, traditional community detection algorithms have limitations when applied to microblogging networks. However, community detection algorithms based on probabilistic topic models are not only applicable to large-scale networks, but can also integrate the semantic information of nodes in microblogging networks with the microblogging network topology. The main probabilistic topic models include LDA (latent Dirichlet allocation) and ATM (author topic model). Summary of the Invention

[0004] The purpose of this invention is to propose a microblog group identification method based on community discovery in order to solve the above problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A microblog group identification method based on community discovery includes the following steps:

[0007] S1. Data collection and cleaning: Extract the relationships between users from blog posts with @ relationships to form a user graph; perform word segmentation on the Chinese corpus; and use word segmentation tools to perform stop word removal, data cleaning, and Chinese word segmentation on the Weibo data to complete data preprocessing.

[0008] S2. Feature extraction and representation: Construct feature indicators for specific groups. By collecting publicly available content from Weibo, relevant features are extracted from attributes such as content and behavior. Based on the Sina Weibo dataset, training and test sets are constructed through manual labeling.

[0009] S3. Establish a classification model and use the optimized Dynamic Topic Model (DTM) to mine specific groups in the Weibo community. Map the Weibo network to a directed weighted network and combine it with the Label Propagation Algorithm (LPA) to discover communities and identify the inherent community structure in the social relationship network.

[0010] S4, Community Tag and Influence Analysis, calculates the user's influence value by summing and weighting the influence of the user's fans, reflecting the influence relationship of the community group.

[0011] Preferably, the identification method is based on the Label Propagation Algorithm (LPA) of a Directed Weighted Network for community detection, and the algorithm is as follows:

[0012] Input: Directed network G(V,E), weight parameters d, total number of nodes in the network n;

[0013] Output: Community C(V,E);

[0014] Specifically, it consists of the following four steps:

[0015] a) Initialize all nodes in network G The labels are assigned to each node in turn. The number of iterations is t=1;

[0016] b) Randomly arrange the nodes in the network to generate sequence X;

[0017] c) For each node, follow the order in X. Update tags, by the person who owns the tags. nodes Neighboring nodes constitute a community set If there are multiple labels, one will be selected randomly.

[0018] d) If the community gathers Each node has a label If all the labels are the most frequently occurring labels among its neighboring nodes, then the community is considered to have reached a stable state, the iteration ends, and the community is obtained. Otherwise, let the iteration number t = t + 1 and go to step b.

[0019] Preferably, the user influence value is calculated using the following formula:

[0020]

[0021] Where INF is the user's influence value, S is the user's attribute characteristic value, B is the user's behavior value, j is the user's fan set, and d is the coefficient (0, 1).

[0022] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0023] 1. This application utilizes an optimized dynamic topic model (DTM) to mine specific groups within the Weibo community. It selects Weibo posts from the past year as the research object, using the topic similarity of posts from different authors as the weight of links between authors, mapping the Weibo network into a directed weighted network. Combined with the label propagation algorithm (LPA), it performs community discovery, identifying the inherent community structure within the social relationship network. This invention conducts in-depth analysis of user relationships in the Weibo network, and based on identification methods for user-generated content characteristics, user association characteristics, and environmental characteristics, it mines potential topics, identifying users with similar interests and active user groups in specific fields. Attached Figure Description

[0024] Figure 1 A schematic diagram of the process structure of a microblog group identification method based on community discovery according to an embodiment of the present invention is shown; Detailed Implementation

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] Please see Figure 1 , the present invention provides a technical solution:

[0027] A microblog group identification method based on community discovery includes the following steps:

[0028] S1. Data Collection and Cleaning: Select Weibo content and Weibo users as data collection objects. Use a web crawler based on the API interface of Sina Weibo Open Platform to acquire data. Collect public Weibo information and user information from Sina Weibo from 2022 to the present as a dataset. Extract the relationships between users from the posts with @ relationships to form a user graph. Perform word segmentation on the Chinese corpus. Use word segmentation tools to perform stop word removal, data cleaning and Chinese word segmentation on the Weibo data to complete data preprocessing.

[0029] S2. Feature Extraction and Representation: Construct feature indicators for a specific group. By collecting publicly available Weibo content, extract relevant features from attributes such as content and behavior. Based on the Sina Weibo dataset, construct training and test sets using manual labeling, and represent them as specific vector forms. Key feature values ​​are extracted as shown in Table 1.

[0030] Table 1. Key features of Weibo users extracted

[0031] Feature categories Feature Description Content Features The following attributes are considered: number of blog post titles, number of blog post topics, percentage of blog posts with URLs, percentage of blog posts with titles, percentage of blog posts with topics, number of words in blog posts, percentage of blog posts containing both topics and titles, percentage of blog posts containing both titles and URLs, percentage of blog posts containing both topics and URLs, percentage of reposted blog posts with URLs, percentage of reposted blog posts with topics, percentage of reposted blog posts with titles, percentage of blog posts containing emoticons, and blog post length. The maximum, minimum, mean, and median values ​​for all of these attributes are also considered. Behavioral characteristics Number of followers, number of followers, follower-to-follower ratio, number of posts, percentage of posts actively mentioned by users, percentage of posts replied to, percentage of posts forwarded, average posting interval, percentage of posts posted during working hours, number of posts per day, number of posts per day with URL. The maximum, minimum, mean, and median values ​​for all of the above attributes.

[0032] S3. Establish a classification model and use the optimized dynamic topic model (DTM) to mine specific groups in the Weibo community. Map the Weibo network to a directed weighted network and combine it with the label propagation algorithm (LPA) to discover communities and identify the inherent community structure in the social relationship network.

[0033] S4, Community Tag and Influence Analysis, calculates the user's influence value by summing and weighting the influence of the user's fans, reflecting the influence relationship of the community group.

[0034] Specifically, such as Figure 1 As shown, the identification method is based on the label propagation algorithm LPA of directed weighted networks for community detection. The algorithm is as follows:

[0035] Input: Directed network G(V,E), weight parameters d, total number of nodes in the network n;

[0036] Output: Community C(V,E);

[0037] Specifically, it consists of the following four steps:

[0038] a) Initialize all nodes in network G The labels are assigned to each node in turn. The number of iterations is t=1;

[0039] b) Randomly arrange the nodes in the network to generate sequence X;

[0040] c) For each node, follow the order in X. Update tags, by the person who owns the tags. nodes Neighboring nodes constitute a community set If there are multiple labels, one will be selected randomly.

[0041] d) If the community gathers Each node has a label If all the labels are the most frequently occurring labels among its neighboring nodes, then the community is considered to have reached a stable state, the iteration ends, and the community is obtained. Otherwise, let the iteration number t = t + 1 and go to step b.

[0042] User influence score is calculated using the following formula:

[0043]

[0044] Where INF is the user's influence value, S is the user's attribute characteristic value, B is the user's behavior value, j is the user's fan set, and d is the coefficient (0, 1).

[0045] This application conducts an in-depth analysis of user relationships on the microblogging network. Based on the identification methods of user-generated content characteristics, user relationship characteristics, and environmental characteristics, it mines potential themes and identifies users with similar interests as well as active user groups in specific fields.

[0046] This application utilizes an optimized Dynamic Topic Model (DTM) to mine specific groups within the Weibo community. It selects Weibo posts from the past year as the research object, using the topic similarity of posts from different authors as the weight of links between authors, mapping the Weibo network into a directed weighted network. Combined with the Label Propagation Algorithm (LPA), it performs community discovery, identifying the inherent community structure within the social relationship network. This invention conducts an in-depth analysis of user relationships in the Weibo network, using identification methods based on user-generated content characteristics, user association characteristics, and environmental characteristics to mine potential topics and identify users with similar interests and active user groups in specific fields.

[0047] The above description of the embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A microblog group identification method based on community discovery, characterized in that, Includes the following steps: S1. Data collection and cleaning: Extract the relationships between users from blog posts with @ relationships to form a user graph; perform word segmentation on the Chinese corpus; and use word segmentation tools to perform stop word removal, data cleaning, and Chinese word segmentation on the Weibo data to complete data preprocessing. S2. Feature extraction and representation: Construct feature indicators for specific groups. By collecting publicly available content from Weibo, relevant features are extracted from attributes such as content and behavior. Based on the Sina Weibo dataset, training and test sets are constructed through manual labeling. S3. Establish a classification model and use the optimized Dynamic Topic Model (DTM) to mine specific groups in the Weibo community. Map the Weibo network to a directed weighted network and combine it with the Label Propagation Algorithm (LPA) to discover communities and identify the inherent community structure in the social relationship network. S4, Community Tag and Influence Analysis, calculates the user's influence value by summing and weighting the influence of the user's fans, reflecting the influence relationship of the community group; The identification method is based on the Label Propagation Algorithm (LPA) of Directed Weighted Networks for community detection, and the algorithm is as follows: Input: Directed network G(V,E), weight parameters d, total number of nodes in the network n; Output: Community C(V,E); Specifically, it consists of the following four steps: a) Initialize all nodes in network G The labels are assigned to each node in turn. The number of iterations is t=1; b) Randomly arrange the nodes in the network to generate sequence X; c) For each node, follow the order in X. Update tags, by the person who owns the tags. nodes Neighboring nodes constitute a community set If there are multiple labels, one will be selected randomly. d) If the community gathers Each node has a label If all the labels are the most frequently occurring labels among its neighboring nodes, then the community is considered to have reached a stable state, the iteration ends, and the community is obtained. Otherwise, let the iteration count t = t + 1 and go to step b; The user influence value is calculated using the following formula: Where INF is the user's influence value, S is the user's attribute feature value, B is the user's behavior value, j is the user's fan set, and d is the coefficient (0, 1).

Images