CN114896522A - Multi-platform information epidemic situation risk assessment method and device - Google Patents

Abstract

The invention discloses a multi-platform information epidemic situation risk assessment method, which comprises the following steps: the method comprises the following steps: collecting flow data of each platform; step two: extracting a domain name list for each piece of stream data to obtain domain name redirection historical information, and matching the domain name redirection historical information with a domain name reliability corpus to obtain a reliability label of the stream data; step three: analyzing the user-defined position of each piece of streaming data to obtain geographic information; step four: grouping the stream data according to two dimensions of geographic information and time; step five: quantifying a static information epidemic situation risk index value based on the number of fans of the user and the reliability label for each group of flow data; step six: and quantifying the dynamic information epidemic situation risk index values for each group of stream data based on the praise number, the forwarding number, the comment number and the reliability label. The invention also provides an evaluation device. The invention reflects the upper risk limit and the information epidemic degree of the information epidemic by constructing the static information epidemic risk index and the dynamic information epidemic risk index.

Description

Multi-platform information epidemic risk assessment method and device

technical field

The invention relates to the technical field of data mining. More specifically, the present invention relates to a multi-platform information epidemic risk assessment method and device.

Background technique

The sudden outbreak of COVID-19 has brought a huge impact on the world. This shock not only involves physical space, but also spreads to cyberspace, with the emergence of an “information epidemic” characterized by the proliferation of fake news. The information epidemic not only seriously interferes with epidemic prevention and control, but also threatens regional security and stability. If the information epidemic risk of each region can be assessed in real time, it will be of great help for online and offline epidemic prevention and control. Therefore, there is an urgent need to design an effective information epidemic risk assessment method and device.

SUMMARY OF THE INVENTION

One object of the present invention is to provide a multi-platform information epidemic risk assessment method and device, which can reflect the information epidemic risk upper limit and information epidemic degree by constructing static information epidemic risk indicators and dynamic information epidemic risk indicators.

In order to achieve these objects and other advantages of the present invention, according to one aspect of the present invention, the present invention provides a multi-platform information epidemic risk assessment method, including: step 1: collecting flow data of each platform; step 2: analyzing each flow data , extract the domain name list, obtain the domain name redirection history information, and match it with the domain name reliability corpus to obtain the reliability label of the flow data; Step 3: For each piece of flow data, analyze the user-defined location to obtain geographic information; Step 4, The flow data is grouped according to the two dimensions of geographic information and time; step 5, for each group of flow data, based on the number of users and reliability labels, quantify the static information that reflects the epidemic risk level of static information in a region within a certain period of time. Epidemic risk index value; Step 6: For each group of streaming data, based on the number of likes, retweets, comments, and reliability labels, quantify the dynamic information epidemic risk that reflects the degree of dynamic information epidemic risk in a region within a certain period of time Index value.

Further, it also includes: collecting domain name reliability labeling data from multiple sources; constructing multi-type reliability labels, assigning unreliability scores, and mapping domain name reliability labeling data from various sources to reliability labels; The reliability label combines the domain name reliability labeling data from various sources to form a domain name reliability corpus. For the domain name reliability labeling data with multiple reliability labels, the reliability label with the lowest unreliability score is used.

Further, formula 1 is used to calculate the static information epidemic risk index value staticIRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all stream data with reliability labels in region c during time period d, fans _i represents the number of followers of the poster of stream data i, r _i represents the unreliability score of stream data i.

Further, formula 2 is used to calculate the dynamic information epidemic risk index value dynamicIRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in the region c during the d time period, like _i represents the number of likes of stream data _i , and ri represents stream data Unreliability score for data i.

Further, a regular expression is used to extract the domain name, and an asynchronous crawler is used to obtain the domain name redirection history information.

Further, the user-defined location is analyzed by using the geocoding service provided by the geographic service provider to obtain geographic information.

Further, the granularity of time grouping is day, week or month, and the granularity of geographic information grouping is city, province or country.

Further, it also includes: obtaining the static information epidemic risk index value and dynamic information epidemic risk index value corresponding to each time period in the historical time, taking the static information epidemic risk index value as the input, and taking the dynamic information epidemic risk index value as the output, Training to obtain a neural network prediction model; input the static information epidemic risk index value of the current time period into the neural network prediction model, output the predicted value of the dynamic information epidemic risk index value, and compare the predicted value with the dynamic information epidemic risk index value of the current time period. By comparison, if the error exceeds the first predetermined range, the dynamic information epidemic risk index value of the current time period is marked, and if the error exceeds the second predetermined range, the flow data of the current time period is collected again.

According to another aspect of the present invention, a multi-platform information epidemic risk assessment device is also provided, including: a collection module for collecting flow data of each platform; a label assigning module for extracting a list of domain names for each piece of flow data, and obtaining Domain name redirection history information, and match with the domain name reliability corpus, and assign reliability labels; geographic information analysis module, used to parse user-defined location for each stream data, and obtain geographic information; grouping module, used to classify according to geographic information The flow data is grouped in the two dimensions of time and time; the static information epidemic risk index value calculation module is used to quantify the static information epidemic in a region within a certain period of time based on the number of users and reliability labels for each group of flow data. The static information epidemic risk index value of the risk level; the dynamic information epidemic risk index value calculation module is used to quantify the number of likes, retweets, comments, and reliability labels for each group of stream data to reflect that an area is in a certain area. The dynamic information epidemic risk index value of the dynamic information epidemic risk degree within the time period.

Further, the static information epidemic risk index value calculation module adopts formula 1 to calculate the static information epidemic risk index value staticIRI _{c, d} :

The dynamic information epidemic risk index value calculation module uses formula 2 to calculate the dynamic information epidemic risk index value dynamicIRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in region c during the d time period, like _i represents the number of likes of stream data i, and fans _i represents stream data The number of followers that the poster of data _i has, and ri represents the unreliability score of stream data i.

The present invention includes at least the following beneficial effects:

The invention comprehensively considers the commonality of multiple platforms, and constructs a static information epidemic risk index and a dynamic information epidemic risk index to comprehensively evaluate the information epidemic risk degree of each platform. The static index reflects the upper limit of the risk of the information epidemic, and the dynamic index evaluates the current information epidemic degree of each platform; the information epidemic risk calculation method based on the redirection technology of the present invention is more accurate than the previous traditional simple matching information epidemic calculation method. comprehensive.

Other advantages, objects, and features of the present invention will appear in part from the description that follows, and in part will be appreciated by those skilled in the art from the study and practice of the invention.

Description of drawings

FIG. 1 is a frame diagram of the present invention.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings, so that those skilled in the art can implement it with reference to the description.

It should be understood that terms such as "having", "comprising" and "including" as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

As shown in FIG. 1 , the embodiment of the present application provides a multi-platform information epidemic risk assessment method, including: Step 1: Collect streaming data of each platform, the platforms include Zhihu, Weibo, twitter, Facebook, telegram, etc. Stream data at least include the text content, release time, user-defined location, number of followers, number of likes and other information of the stream data;

Optionally, daily stream data collected on social platforms such as Weibo, facebook, twitter, and telegram with keywords such as "COVID-19" and "Virus", the server uses the pull-push scheme in the transport layer ZMQ to The stream data is evenly distributed to each port of the server; after the server receives the stream data, it stores the stream data of each channel in the corresponding table, that is, each channel has its own stream data storage table; the database can use Elasticsearch Distributed storage database to realize fast processing and storage of large-scale data; the data owned by each platform at least include mid, text, time, user_location, user_fans, like_num; mid represents the id of the current stream data, text represents the text content of the current stream data, and time represents The current streaming data publishing time, user_location represents the user-defined location of the current streaming data poster, user_fans represents the number of fans of the current streaming data poster, and like_num represents the current streaming data likes number;

Step 2: For each piece of flow data, extract the domain name list, obtain the domain name redirection history information, and match with the domain name reliability corpus to obtain the reliability label of the flow data; extract the domain name list contained in the text content, after extracting the domain name list. , and further obtain the domain name redirection history information; then, match the domain name redirection history information of the stream data with the domain name reliability corpus to obtain the reliability label of the stream data;

Step 3: For each stream data, analyze the user-defined location to obtain geographic information;

Step 4: Group the streaming data according to the two dimensions of geographic information and time; in the database, group all data according to time and location to ensure that the time and location fields of each group are the same;

Step 5: For each group of streaming data, based on one or more of the number of users and reliability labels, quantify the static information epidemic risk index value (risk Spend);

Step 6: For each set of streaming data, based on one or more of the likes, retweets, comments, and reliability labels, quantify the dynamic information that reflects the epidemic risk level of dynamic information in a region within a certain period of time Epidemic risk index value (risk degree);

It can be seen that, in view of the great differences in the data format and content of different social platforms, this embodiment comprehensively considers the commonality of multiple platforms, and constructs static information epidemic risk indicators and dynamic information epidemic risk indicators to comprehensively evaluate the information epidemic risk of each platform. The static index reflects the upper limit of the information epidemic risk, and the dynamic index evaluates the current information epidemic degree of each platform; the information epidemic risk calculation method based on redirection technology is more accurate than the traditional simple matching information epidemic calculation method. comprehensive.

Aiming at the problems of relying on the domain name reliability corpus and false domain names with "vesticles" in the judgment of false information, a method of integrating the marked data of domain names from different sources is established, and the redirection technology is introduced for domain name matching, which makes the judgment of false information more rigorous and accurate. This indicator system can not only assess the risk of information epidemics, but also deeply explain the mechanism behind the evolution of information epidemics.

In other embodiments, the method further includes: collecting domain name reliability labeling data from multiple sources; constructing multi-type reliability labels, assigning unreliability scores, and mapping domain name reliability labeling data from various sources to reliability labels above; merge the domain name reliability labeling data from various sources according to the reliability label to form a domain name reliability corpus. For the domain name reliability labeling data with multiple reliability labels, the one with the lowest unreliability score is used as the reliability label;

Optionally, first collect authoritative domain name reliability labeling data widely used in academia, where domain name reliability labeling data includes mediabiasfactcheck, decodex, etc., and its main format is a domain name-reliability label key-value pair, such as {“79Days.News ”: conspiracy theory}, {“lung.ory”: science}; this is just an example, but not limited to the above two domain name reliability annotation data, the domain name reliability annotation data that conforms to this category are all within the scope of protection;

Construct a unified reliability label, including eight categories of "science", "mainstream media", "sarcasm", "title party", "other", "politics", "false information", "pseudoscience", and its unreliability Scores range from 1 to 8, in order of unreliability from low to high. Among them, the labels of the domain name reliability marking data from different sources are not necessarily the same, and they need to be unified to facilitate merging. Map the domain name reliability mark data labels of various sources to a unified reliability label, such as mapping the "conspiracy theory" label in {"79Days.News": conspiracy theory} to the "pseudoscience" label;

Then merge the domain name reliability data from various sources. For domain names with multiple labels, the label with the lowest reliability is used as the final label; for example, the label of 79Days.News in mediabiasfactcheck is "false information", and in decodex, it is "pseudoscience" ”, the domain name will be labeled as “pseudoscience” after the final merger;

Aiming at the problem that reliability evaluation relies heavily on the domain name reliability corpus, this embodiment comprehensively considers the different characteristics of mainstream websites in various regions, and based on the current academic research on the reliability of streaming data, establishes a combination of false domain name tagging data from different sources. Method; This method can capture more comprehensive false information by merging false domain name tagging data from various sources, thereby providing a more in-depth and comprehensive explanation of the mechanism behind the evolution of information epidemics.

In other embodiments, formula 1 is used to calculate the static information epidemic risk index value staticIRI _{c, d} :

表示发帖数量。Among them, c represents a region, d represents a time period, T _{c, d} represents all stream data with reliability labels in region c during time period d, fans _i represents the number of followers of the poster of stream data i, r _i represents the unreliability score of stream data i,

Indicates the number of posts.

In other embodiments, the dynamic information epidemic risk index value dynamicIRI _{c, d} is calculated by formula 2:

表示发帖数量。Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in the region c during the d time period, like _i represents the number of likes of stream data _i , and ri represents stream data Unreliability score for data i,

Indicates the number of posts.

In other embodiments, a regular expression is used to extract the domain name, and an asynchronous crawler is used to obtain the domain name redirection history information; optionally, for each piece of flow data, a regular expression is used to match all domain names in the text field of the flow data. Then, use asynchttp to redirect these domain names to obtain the redirection history of these domain names; then use regular expressions to match the redirection history of these domain names with the domain name reliability corpus to obtain the reliability annotation of the domain name redirection history; The lowest reliability is used as the reliability label of the stream data; the reliability label of the stream data is stored in the database as the reliability field; this embodiment breaks through the traditional method of evaluating the reliability of stream data only by simple matching Considering the possible existence of “domain name vests” in false information, the real domain name of false information is obtained by using redirection technology, so as to evaluate the risk of information epidemic more comprehensively and accurately; therefore, information epidemic based on redirection technology Compared with the traditional simple matching information epidemic calculation method, the risk calculation method has more accurate and comprehensive results.

In other embodiments, the geocoding service provided by the geographic service provider is used to parse the user-defined location to obtain geographic information; optionally, for each stream data, extract its user_location field, and use requests to construct a request and send it to ArcGIS The server obtains the structured geographic information of user_location, and stores the structured geographic information as a location field in the database; for example, if user_location is "Beijing University of Aeronautics and Astronautics", after sending it to ArcGIS, the geographic information of "China, Beijing" is obtained.

In other embodiments, the granularity of time grouping is day, week or month, and the granularity of geographic information grouping is city, province or country; optionally, in the database, all data are grouped according to time and location to ensure that each The time and location fields of the group are the same. Among them, the granularity of time can be set to days, weeks, and months, and the granularity of location can be set to provinces and countries. The specific granularity can be adjusted according to the analysis needs.

In other embodiments, the method further includes: obtaining the static information epidemic risk index value and the dynamic information epidemic risk index value corresponding to each time period in the historical time, using the static information epidemic risk index value as input, and using the dynamic information epidemic risk index as input The value is the output, and the neural network prediction model is obtained by training; the static information epidemic risk index value of the current time period is input into the neural network prediction model, the predicted value of the dynamic information epidemic risk index value is output, and the predicted value is compared with the dynamic information epidemic situation of the current time period. The risk index values are compared, and if the error exceeds the first predetermined range, the dynamic information epidemic risk index value of the current time period is marked, and if the error exceeds the second predetermined range, the flow data of the current time period is collected again;

In the above embodiment, within a period of time, the static information epidemic risk index value has a strong correlation with the dynamic information epidemic risk index value, but the static information epidemic risk index value is less affected by human beings, and the dynamic information epidemic risk index value is relatively small. It is greatly affected by human beings and is easy to be manipulated by humans. The purpose of this embodiment is to perform internal detection on the static information epidemic risk index value and dynamic information epidemic risk index; Network prediction model; when the static information epidemic risk index value of the current time period is calculated, it is input into the neural network prediction model, the predicted value of the dynamic information epidemic risk index value is output, and the predicted value is compared with the dynamic information epidemic calculated by formula 2. Compare the actual value of the risk index and calculate the error; compare the error with the first predetermined range and the second predetermined range, the first predetermined range can be selected as 20-30%, when the error between the predicted value and the actual value is greater than the first predetermined range, Indicates that the actual value of the dynamic information epidemic risk index is high to a certain extent, and the actual value of the dynamic information epidemic risk index at this time is marked for users to distinguish; when the error between the predicted value and the actual value is greater than the second predetermined range, the first predetermined range can be The selection of 50-60% indicates that the actual value of the dynamic information epidemic risk index is extremely high, and the actual value of the dynamic information epidemic risk index at this time is inaccurate, and data collection needs to be re-calculated.

The embodiment of the present application also provides a multi-platform information epidemic risk assessment device, including: a collection module for collecting flow data of each platform; a label assigning module for extracting a domain name list and obtaining domain name redirection for each piece of flow data Historical information, matched with the domain name reliability corpus, and assigned reliability labels; the geographic information parsing module is used to parse the user-defined location for each stream data to obtain geographic information; the grouping module is used to analyze the geographic information and time The flow data is grouped by one dimension; the static information epidemic risk index value calculation module is used for each group of flow data, based on the number of users and the reliability label, to quantify and obtain the static information epidemic risk degree in a region within a certain period of time. Static information epidemic risk index value; dynamic information epidemic risk index value calculation module, used for each group of stream data, based on the number of likes, retweets, comments, and reliability labels to quantify and reflect a region within a certain period of time. The dynamic information epidemic risk index value of the dynamic information epidemic risk degree;

In this embodiment, a processor and a memory are used to realize the establishment of a collection module, a label assignment module, a geographic information analysis module, a grouping module, a static information epidemic risk index value calculation module, and a dynamic information epidemic risk index value calculation module, so as to realize many of the above embodiments. The platform information epidemic risk assessment method, please refer to the description above for details; the information epidemic risk degree can be calculated every day, and on this basis, the dynamic and static two-week information epidemic risk indicators can be obtained; all these indicators are stored in the database by time and region Therefore, by querying the database, you can obtain the evolution trend of the information epidemic in various countries and provinces from the monitoring date to the current date, analyze the mechanism of the evolution trend, and further predict the future evolution of the information epidemic; Early warning information, quickly grasp the development and change trends of the information epidemic, better understand and guide the mentality of netizens, and play the positive role of online public opinion, which will help the real epidemic situation offline.

In other embodiments, the static information epidemic risk index value calculation module uses Formula 1 to calculate the static information epidemic risk index value staticIRI _c,d :

The dynamic information epidemic risk index value calculation module uses formula 2 to calculate the dynamic information epidemic risk index value dynamicIRI _{c, d} :

表示发帖数量。Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in region c during the d time period, like _i represents the number of likes of stream data i, and fans _i represents stream data The number of followers of the poster of data _i , ri represents the unreliability score of stream data i,

Indicates the number of posts.

The number of apparatuses and processing scales described here are intended to simplify the description of the present invention. The application, modification and variation of the multi-platform information epidemic risk assessment method of the present invention will be obvious to those skilled in the art.

Although the embodiment of the present invention has been disclosed as above, it is not limited to the application listed in the description and the embodiment, and it can be applied to various fields suitable for the present invention. For those skilled in the art, it can be easily Therefore, the invention is not limited to the specific details and illustrations shown and described herein without departing from the general concept defined by the appended claims and the scope of equivalents.

Claims (10)

1. A multi-platform information epidemic risk assessment method, characterized in that it includes: Step 1: Collect the streaming data of each platform; Step 2: For each stream data, extract the domain name list, obtain the domain name redirection history information, and match with the domain name reliability corpus to obtain the reliability label of the stream data; Step 3: For each stream data, analyze the user-defined location to obtain geographic information; Step 4: Group the streaming data according to the two dimensions of geographic information and time; Step 5: For each set of streaming data, based on the number of users and reliability labels, quantify the static information epidemic risk index value that reflects the static information epidemic risk level of a region in a certain period of time; Step 6: For each set of streaming data, based on the number of likes, retweets, comments, and reliability labels, quantify the dynamic information epidemic risk index value that reflects the degree of dynamic information epidemic risk in a region within a certain period of time. 2. The multi-platform information epidemic risk assessment method as claimed in claim 1, further comprising: Collect domain name reliability annotation data from multiple sources; Build multi-class reliability labels, assign unreliability scores, and map domain name reliability labeling data from various sources to reliability labels; According to the reliability labels, the domain name reliability labeling data from various sources is combined to form a domain name reliability corpus. For the domain name reliability labeling data with multiple reliability labels, the reliability label with the lowest unreliability score is used. 3. multi-platform information epidemic risk assessment method as claimed in claim 1, is characterized in that, adopts formula 1 to calculate static information epidemic risk index value staticJRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all stream data with reliability labels in region c during time period d, fans _i represents the number of followers of the poster of stream data i, r _i represents the unreliability score of stream data i. 4. multi-platform information epidemic risk assessment method as claimed in claim 1, is characterized in that, adopts formula 2 to calculate dynamic information epidemic risk index value dynamicIRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in the region c during the d time period, like _i represents the number of likes of stream data _i , and ri represents stream data Unreliability score for data i. 5. The multi-platform information epidemic risk assessment method according to claim 1, characterized in that, a regular expression is used to extract the domain name, and an asynchronous crawler is used to obtain the domain name redirection history information. 6 . The multi-platform information epidemic risk assessment method according to claim 1 , wherein the geographic information is obtained by analyzing the user-defined location by using a geocoding service provided by a geographic service provider. 7 . 7. The multi-platform information epidemic risk assessment method according to claim 1, wherein the granularity of time grouping is day, week or month, and the granularity of geographic information grouping is city, province or country. 8. The multi-platform information epidemic risk assessment method of claim 1, further comprising: Obtain the static information epidemic risk index value and dynamic information epidemic risk index value corresponding to each time period in the historical time, take the static information epidemic risk index value as the input, and take the dynamic information epidemic risk index value as the output, and train to obtain the neural network prediction model ; Input the static information epidemic risk index value of the current time period into the neural network prediction model, output the predicted value of the dynamic information epidemic risk index value, and compare the predicted value with the dynamic information epidemic risk index value of the current time period. If the error exceeds the second predetermined range, the dynamic information epidemic risk index value of the current time period is marked. If the error exceeds the second predetermined range, the flow data of the current time period is collected again. 9. The multi-platform information epidemic risk assessment device according to claim 1, characterized in that, comprising: The collection module is used to collect the flow data of each platform; The label assignment module is used to extract the domain name list for each stream data, obtain the domain name redirection history information, match with the domain name reliability corpus, and assign the reliability label; The geographic information analysis module is used to analyze the user-defined location for each stream data to obtain geographic information; The grouping module is used to group the streaming data according to the two dimensions of geographic information and time; The static information epidemic risk index value calculation module is used to quantify the static information epidemic risk index value that reflects the static information epidemic risk level of a region in a certain period of time based on the number of users and reliability labels for each group of streaming data; The dynamic information epidemic risk index value calculation module is used to quantify the dynamic information reflecting the epidemic risk degree of dynamic information in a region within a certain period of time based on the number of likes, retweets, comments, and reliability labels for each group of stream data. Information epidemic risk index value. 10. The multi-platform information epidemic risk assessment device according to claim 1, wherein the static information epidemic risk index value calculation module adopts formula 1 to calculate the static information epidemic risk index value staticIRI _{c, d} :

The dynamic information epidemic risk index value calculation module uses formula 2 to calculate the dynamic information epidemic risk index value dynamicIRI _{c, d} :

Among them, c represents a region, d represents a time period, T _{c, d} represents all the stream data with reliability labels in region c during the d time period, like _i represents the number of likes of stream data i, and fans _i represents stream data The number of followers that the poster of data _i has, and ri represents the unreliability score of stream data i.

Images