Micro-blog text-based subway emergency network public opinion analysis method and device

Abstract

The application provides a subway emergency network public opinion analysis method and device based on micro-blog texts, relates to the field of emergency management, and comprises the following steps: formulating a network public opinion collection rule, and determining target network public opinion data of network public opinion information in a subway emergency according to the network public opinion collection rule, wherein the network public opinion collection rule comprises a collection time rule, a collection keyword rule and a collection field rule; preprocessing the target network public opinion data, and extracting a feature vector of text in the target network public opinion data; respectively performing time sequence analysis, spatial analysis, sentiment analysis and topic and people's livelihood analysis on the target network public opinion data after preprocessing, statistically processing each analysis data, and obtaining statistical data; and formulating an optimal emergency rescue scheme according to the statistical data. The application analyzes the subway emergency from the aspects of time, space, sentiment and topic, and provides support for safe operation of a subway system, analysis of network public opinions and guidance of public opinions.

Description

Technical Field

[0001] This application relates to the field of emergency management, and in particular to a method and device for analyzing online public opinion on subway emergencies based on Weibo text. Background Technology

[0002] With the continuous development and progress of society, subways have become an important means of transportation. The coupling of various emergencies with sudden surges in passenger flow poses significant risks to the safe operation of the subway system. How to promptly and effectively grasp the public's emotional tendencies and needs in the face of subway emergencies has become a key focus and challenge in current subway emergency management. Meanwhile, with the popularization of internet technology, more and more people are willing to express their thoughts and needs on social media platforms. Therefore, online public opinion analysis brings new approaches and challenges to subway emergency management.

[0003] Currently, online public opinion analysis in the field of emergency management mainly focuses on the disaster itself. In the emergency management of subways, there is a lack of experience in online public opinion analysis and public opinion guidance, and there is still considerable room for improvement. Summary of the Invention

[0004] To address the aforementioned issues, a method and device for analyzing online public opinion regarding subway emergencies based on microblog text is proposed. This method utilizes web crawling technology to define keywords for crawling online public opinion related to subway emergencies, identifying key information such as publication time, location, and text content. It then analyzes online public opinion regarding subway emergencies from the perspectives of time, space, sentiment, and trending topics to determine the most needed emergency assistance for the public at different time periods. This helps subway emergency management departments to take timely emergency measures, ensuring the public's basic needs for food, clothing, shelter, and transportation, and providing support for the safe operation of the subway system, public opinion analysis, and public opinion guidance.

[0005] The first aspect of this application proposes a method for analyzing online public opinion regarding subway emergencies based on microblog text, including:

[0006] Formulate rules for collecting online public opinion, and determine the target online public opinion data for subway emergencies based on the rules. The rules for collecting online public opinion include rules for collection time, rules for collection keywords, and rules for collection fields.

[0007] The target online public opinion data is preprocessed to extract feature vectors of the text within the target online public opinion data;

[0008] The preprocessed target online public opinion data is subjected to time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis, and statistical data is obtained by statistical analysis of each analysis data.

[0009] Based on the statistical data, develop the best emergency response plan.

[0010] Optionally, the data collection time rules include:

[0011] Determine the analysis requirements for the aforementioned subway emergencies;

[0012] If the analysis requirement is to analyze historical emergencies, the data collection time range is selected from one week before the occurrence of the historical emergencies to one week after the end of the historical emergencies.

[0013] If the analysis requirement is real-time monitoring, the online public opinion information is collected in real time, and the collection method is to collect once every Δt minutes, with each collection containing the online public opinion information from the previous Δt minutes.

[0014] Optionally, the keyword collection rules include:

[0015] The number of required words and the number of arbitrarily included words in the online public opinion information text are determined, wherein the number of required words is greater than 0, and the number of arbitrarily included words is not limited.

[0016] Optionally, the collection field rules include:

[0017] According to the rules of the social media platform, target fields of online public opinion information in the subway emergency are extracted. The target fields include the poster's nickname, the poster's authentication information, the posting time, the posting content, the number of reposts, the number of comments, and the number of likes.

[0018] Optionally, the preprocessing of the target online public opinion data to extract feature vectors of the text within the target online public opinion data includes:

[0019] The target online public opinion data is processed by Chinese word segmentation to determine a vocabulary list;

[0020] Iterate through each word in the vocabulary list. If the text length of a word is less than 2 or a word is in the stop word list, then remove that word.

[0021] The TF-IDF weights of the remaining words after the removal process are calculated. The steps for calculating the TF-IDF weights of any word are as follows:

[0022]

[0023] Where n is the number of times the word appears in the public opinion text, N is the number of words remaining in the public opinion text after removing stop words, M is the number of public opinion texts, and m is the number of public opinion texts containing the word.

[0024] The feature weights are constructed based on the TF-IDF weights and the remaining vocabulary, and are formulated as follows:

[0025] Y i ={w1:TF_IDF(w1),w2:TF_IDF(w2),…,w m :TF_IDF(w m )};

[0026] For a target network public opinion data, the remaining words after the removal process are sorted from largest to smallest according to the TF-IDF weights, and the words and TF-IDF weights of the first Vim position are retained to complete the dimensionality reduction of the feature vector.

[0027] Optionally, the step of performing time series analysis on the preprocessed target online public opinion data includes:

[0028] Based on the preprocessed target online public opinion data, determine the number of online public opinion events in each time period and plot the time-quantity change curve;

[0029] By analyzing the time-quantity change curve, the likelihood of an online public opinion outbreak can be determined.

[0030] Optionally, the spatial analysis of the preprocessed target online public opinion data includes:

[0031] Obtain the geographic location information of the preprocessed target network public opinion data, wherein the geographic location information includes subway line information, subway station information and subway section information;

[0032] By combining the text content of the preprocessed target network public opinion data, the geographical location information is identified to determine distress messages and refuge information;

[0033] The number of distress messages and evacuation messages in the geographical location information is counted, and the number and location of the information are plotted on the subway network map using a heat map method.

[0034] Optionally, the step of performing sentiment analysis on the preprocessed target online public opinion data includes:

[0035] The sentiment value of the target network public opinion data is calculated based on the Naive Bayes algorithm;

[0036] The target online public opinion data is divided into sentiment values ​​according to preset rules to determine sentiment tendencies, wherein the sentiment tendencies include negative emotions, neutral emotions and positive emotions.

[0037] The preprocessed target network public opinion data is divided according to a preset time interval, and the sentiment tendency of each time group after division is calculated. If the sentiment tendency of a certain time group is negative or the difference between the sentiment value of a certain time group and the sentiment value of the previous time group is not less than a preset value, the subway public opinion management department will release an official Weibo post.

[0038] The sentiment value of the target network public opinion data is determined based on the geographical location information. If the sentiment value of a station or section is lower than a preset value, the reasons for the sentiment value of the station or section are analyzed, and emergency rescue measures are taken for the station or section.

[0039] Optionally, the step of performing topic and public opinion analysis on the preprocessed target online public opinion data includes:

[0040] Based on the public safety triangle theory model, the topic is divided into disaster environment, disaster impact, emergency management, disaster-bearing carriers, positive evaluation and negative evaluation, and people's livelihood is divided into clothing, food, housing and transportation.

[0041] Based on the feature vectors, the preprocessed target online public opinion data is divided into corresponding topic categories and livelihood categories;

[0042] The number of public opinions for each of the aforementioned topic categories and livelihood categories within each time period was statistically analyzed, and a time-quantity change curve was plotted.

[0043] Based on the plotted time-quantity change curve, the public's focus and needs for each period can be obtained.

[0044] The second aspect of this application proposes a device for analyzing online public opinion regarding subway emergencies based on microblog text, comprising:

[0045] The acquisition module is used to formulate online public opinion collection rules and determine the target online public opinion data in the subway emergency based on the online public opinion collection rules. The online public opinion collection rules include collection time rules, collection keyword rules and collection field rules.

[0046] The preprocessing module is used to preprocess the target online public opinion data and extract the feature vectors of the text within the target online public opinion data;

[0047] The multi-analysis module is used to perform time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis on the preprocessed target online public opinion data, and to statistically analyze the data from each analysis to obtain statistical data.

[0048] The output module is used to formulate the best emergency rescue plan based on the statistical data.

[0049] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0050] By using web crawler technology, the system specifies the keywords to be crawled for online public opinion regarding subway emergencies, identifies key information such as the release time, location, and text content of online public opinion information, and analyzes online public opinion regarding subway emergencies from the perspectives of time, space, emotion, and hot topics. This allows the system to determine the emergency assistance most needed by the public at different times, thereby helping subway emergency management departments to take timely emergency measures, ensuring the public's basic needs for food, clothing, housing, and transportation, and providing support for the safe operation of the subway system, public opinion analysis, and public opinion guidance.

[0051] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0052] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0053] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present application of a method for analyzing online public opinion regarding subway emergencies based on microblog text;

[0054] Figure 2 This is a flowchart illustrating a time series analysis method for analyzing online public opinion on subway emergencies based on Weibo text, as shown in an exemplary embodiment of this application.

[0055] Figure 3 This is a flowchart illustrating the spatial analysis of a subway emergency network public opinion analysis method based on microblog text, as shown in an exemplary embodiment of this application.

[0056] Figure 4 This is a flowchart illustrating the sentiment analysis method for analyzing online public opinion on subway emergencies based on Weibo text, as shown in an exemplary embodiment of this application.

[0057] Figure 5 This is a flowchart illustrating an exemplary embodiment of the present application of a method for analyzing online public opinion on subway emergencies based on microblog text, focusing on topic and public sentiment analysis.

[0058] Figure 6 This is a block diagram illustrating an exemplary embodiment of the present application of a network public opinion analysis device for subway emergencies based on Weibo text. Detailed Implementation

[0059] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0060] Figure 1 This is a flowchart illustrating an exemplary embodiment of the present application of a method for analyzing online public opinion regarding subway emergencies based on microblog text, comprising:

[0061] Step 101: Formulate online public opinion collection rules, and determine the target online public opinion data for subway emergencies based on the online public opinion collection rules. The online public opinion collection rules include collection time rules, collection keyword rules, and collection field rules.

[0062] In this embodiment of the application, the social media platform used is Sina Weibo, and the rules for collecting online public opinion include three types: time, keywords, and fields.

[0063] The data collection time rules are as follows:

[0064] Determine the analysis requirements for subway emergencies;

[0065] If the analysis requirement is to analyze historical emergencies, the data collection period should be from one week before the historical emergency occurred to one week after the historical emergency ended.

[0066] If the analysis requirement is real-time monitoring, online public opinion information is collected in real time, with the collection method being once every Δt minutes, and each collection includes online public opinion information from the previous Δt minutes.

[0067] Depending on the urgency of the event, Δt takes values ​​of 30, 20, 10, and 5.

[0068] In one possible implementation, information is collected every 5 minutes.

[0069] The rules for collecting keywords are as follows:

[0070] The number of mandatory words and the number of arbitrarily included words in online public opinion information text are determined. The number of mandatory words is greater than 0, and the number of arbitrarily included words is not limited.

[0071] In this application embodiment, the required vocabulary includes subway, Line X, XX station, XX station, and any list of included vocabulary includes words related to emergencies. Typhoon disasters include typhoon, rainstorm and strong wind, and flood disasters include flood, flooding, waterlogging, being submerged, rainstorm and heavy rainfall.

[0072] In one possible implementation, the word "site A" must be included, and the words "heavy rain" and "flooding" can be included arbitrarily.

[0073] The rules for collecting fields are as follows:

[0074] According to the rules of social media platforms, target fields of online public opinion information in subway emergencies are extracted. These target fields include the poster's nickname, the poster's authentication information, the posting time, the posting content, the number of reposts, the number of comments, and the number of likes.

[0075] In this embodiment of the application, according to the rules of the Sina Weibo website, fields can be extracted from the HTML code of the Sina Weibo platform webpage, including: the author's nickname, the author's authentication information, the posting time, the posting content, the number of reposts, the number of comments, and the number of likes.

[0076] In this embodiment of the application, after determining the rules for collecting online public opinion through the above three rules, the online public opinion information is crawled using Python code and saved to a MySQL database.

[0077] Step 102: Preprocess the target online public opinion data and extract the feature vectors of the text within the target online public opinion data.

[0078] In this embodiment of the application, the steps for preprocessing target network public opinion data include: Chinese word segmentation, stop word filtering, calculation of TF-IDF weights, construction of feature vectors, and dimensionality reduction processing, specifically including:

[0079] Perform Chinese word segmentation on the target online public opinion data to determine the vocabulary list.

[0080] In this embodiment of the application, the Jieba library in Python is used for Chinese word segmentation.

[0081] Iterate through each word in the vocabulary list. If the text length of a word is less than 2 or a word is in the stop word list, then remove that word.

[0082] In this embodiment of the application, words with a text length of less than 2 and words located in the stop word list are removed.

[0083] The TF-IDF weights of the remaining words after the removal process are calculated. The steps for calculating the TF-IDF weights of any word are as follows:

[0084]

[0085] Where n is the number of times the word appears in the public opinion text, N is the number of words remaining in the public opinion text after removing stop words, M is the number of public opinion texts, and m is the number of public opinion texts containing the word.

[0086] Feature weights are constructed based on TF-IDF weights and the remaining vocabulary, and the formula is as follows:

[0087] Y i ={w1:TF_IDF(w1),w2:TF_IDF(w2),…,w m :TF_IDF(w , )}.

[0088] In this embodiment of the application, the feature weights of words are constructed based on the TF-IDF weights of the words.

[0089] For a target network public opinion data, the remaining words after the removal process are sorted from largest to smallest according to their TF-IDF weights, and the words and their TF-IDF weights in the first Vim position are retained to complete the dimensionality reduction of the feature vector.

[0090] In this embodiment of the application, the feature vector is subjected to dimensionality reduction processing.

[0091] In one possible implementation, Vim = 50.

[0092] Step 103: Perform time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis on the preprocessed target online public opinion data, and statistically analyze the data from each analysis to obtain statistical data.

[0093] In this embodiment of the application, the preprocessed target online public opinion data is analyzed in four specific aspects.

[0094] The goal of time series analysis is to statistically analyze the quantity of online public opinion in each time period, plot the time-quantity change curve, and analyze the possibility of online public opinion outbreaks by analyzing the trend of changes in the quantity of online public opinion. Specifically, this includes:

[0095] Step 201: Based on the preprocessed target online public opinion data, determine the number of online public opinion events in each time period and plot the time-quantity change curve.

[0096] In this embodiment of the application, the time series analysis of online public opinion can be divided into three types according to the different statistical time precision: daily public opinion time series change analysis, hourly public opinion time series change analysis, and 5-minute public opinion time series change analysis.

[0097] Step 202: By analyzing the time-quantity change curve, determine the likelihood of an outbreak of online public opinion.

[0098] Taking the 5-minute time-series analysis of public opinion changes as an example, we analyze the possibility of an outbreak of online public opinion.

[0099] In one possible implementation, if the number of public opinion posts in the current 5 minutes and the previous 5 minutes are both greater than 100, and the ratio of the number of public opinion posts in the current 5 minutes to the number of public opinion posts in the previous 5 minutes is greater than 2, it indicates that there is a possibility of an online public opinion outbreak. The larger the ratio, the higher the possibility of an outbreak. At this time, the subway emergency management should establish a public opinion working group to monitor the information posted by the public on social media in real time, and release official Weibo information to remind passengers to travel safely and reduce public panic.

[0100] In terms of time analysis, it can identify the occurrence of public opinion events and issue early warnings and guide public opinion.

[0101] The goal of spatial distribution analysis is to identify the stations or sections of highest public concern and, from these, pinpoint severely affected stations and temporary shelters, specifically including:

[0102] Step 301: Obtain the geographic location information of the preprocessed target network public opinion data, including subway line information, subway station information, and subway section information.

[0103] In this embodiment, the text matching rule for subway lines satisfies: Line XX, Line XX, where XX is the subway line name determined by the subway network. A subway line name database is constructed based on the subway network of the city under study, and then this rule is used for matching to obtain the subway line information for each piece of online public opinion data. If no field satisfies the matching rule, the geographical information is empty.

[0104] In one possible embodiment, the subway line is Line 2.

[0105] In this embodiment, the text matching rule for subway stations satisfies: XX station, XX station, where XX is the subway station name determined by the subway network. A subway station name database is constructed based on the subway network of the city under study, and then matched using this rule to obtain the subway station information for each piece of online public opinion data. If no field satisfies the matching rule, the geographical information is empty.

[0106] In one possible embodiment, the subway station is the Civic Center Station.

[0107] In this embodiment, the text matching rule for subway sections satisfies: XX1-XX2 section, XX1 station-XX2 station, XX1-XX2 tunnel, XX1XX2 section, XX1XX2 tunnel, XX1 to XX2, heading to XX2, leaving XX1. Here, XX1 and XX2 are the names of station 1 and station 2, respectively. The station names are determined by the subway network of the city under study. A subway station name database is constructed based on the subway network of the city under study, and then this rule is used for matching to obtain the subway section information for each piece of online public opinion data. If no field satisfies the matching rule, the geographical information is empty.

[0108] In one possible embodiment, the subway section is from Civic Center Station to Garden Road Station.

[0109] Step 302: Combine the text content of the preprocessed target network public opinion data to identify the geographical location information and determine the distress message and refuge message.

[0110] In this embodiment of the application, after obtaining the geographical location of each piece of target online public opinion data, it can be determined whether the information is about distress or evacuation based on the text content of the target online public opinion data. If there are many distress messages near a certain station, it indicates that the station is a severely affected station; if there are many evacuation messages near a certain station, it indicates that there are emergency evacuation sites near the station.

[0111] If any of the following words are present, it indicates a distress message: homeless, flooded, accident, out of service, washed away, interrupted, trapped, missing, out of contact, serious, request, help, distress, need, urgent, emergency; if any of the following words are present, it indicates a refuge message: refuge, temporary place, emergency place.

[0112] Step 303: Count the number of distress and evacuation information in the geographical location information, and plot the number and location of the information on the subway network map using a heat map.

[0113] In this embodiment of the application, the number of distress calls and evacuation requests for each station or section is counted, and ArcGIS technology is used to plot the information on the subway network map using a heat map method.

[0114] In one possible implementation, the number of distress messages is represented in red, indicating severely affected sites; the number of evacuation messages is represented in green, indicating the location of temporary shelters.

[0115] In terms of spatial analysis, it helps to identify severely affected stations, sections of track, and available temporary shelters.

[0116] The goal of online public opinion sentiment analysis is to obtain the public's sentiment scores and sentiment tendencies, specifically including:

[0117] Step 401: Calculate the sentiment value of the target network public opinion data using the Naive Bayes algorithm.

[0118] In this embodiment of the application, the snowNLP dependency library of Python is used to calculate the sentiment value, and the calculated sentiment value is between 0 and 1.

[0119] Step 402: Divide the sentiment values ​​of the target online public opinion data according to preset rules to determine the sentiment tendency, which includes negative sentiment, neutral sentiment and positive sentiment.

[0120] In this embodiment of the application, e is defined as the emotion value.

[0121] If 0 ≤ e ≤ 0.4, then the sentiment expressed in this message is negative.

[0122] If 0.4 < e < 0.6, then the sentiment expressed in this public opinion message is neutral.

[0123] If 0.6≤e≤1, then the sentiment expressed in this public opinion message is positive.

[0124] Step 403: Divide the preprocessed target network public opinion data according to the preset time interval, and calculate the sentiment tendency of each time group after division. If the sentiment tendency of a certain time group is negative or the difference between the sentiment value of a certain time group and the sentiment value of the previous time group is not less than the preset value, the subway public opinion management department will release an official Weibo post.

[0125] In this embodiment of the application, the preset time interval is 5 minutes.

[0126] In one possible implementation, if the public sentiment value for the current time period is lower than 0.4, or if the public sentiment value for the current time period is 0.2 or more lower than the public sentiment value for the previous time period, the subway public opinion management department should promptly release an official Weibo post to announce the actual situation of the emergency response to the subway incident, and reassure the public and reduce their concerns.

[0127] Step 404: Determine the sentiment value of the target network public opinion data based on the geographical location information. If the sentiment value of a station or section is lower than the preset value, analyze the reasons for the sentiment value of the station or section and take emergency rescue measures for the station or section.

[0128] In this embodiment of the application, for each station or section, the sentiment value of the station or section is represented by the average value of all public opinion information sentiment values, and the sentiment value heat map is drawn on the subway network using ArcGIS technology.

[0129] In one possible implementation, for stations with a public sentiment score below 0.4, the reasons for the low public sentiment score are analyzed, emergency rescue measures are taken for the station in a timely manner, and the public sentiment is appeased through official media.

[0130] The goal of online public opinion topic and livelihood analysis is to obtain information on changes in public needs and priorities, specifically including:

[0131] Step 501: Based on the public safety triangle theoretical model, the topic is divided into disaster environment, disaster impact, emergency management, disaster-bearing carrier, positive evaluation and negative evaluation, and people's livelihood is divided into clothing, food, housing and transportation.

[0132] In this application embodiment, disaster environment topics include rainstorms, waterfalls, water accumulation, torrential rain, heavy rainfall, heavy rain, continuous, situation, floods, water levels, rainfall, slippery conditions, typhoons, earthquakes, and strong winds; disaster impact topics include being trapped, being grounded, impact, service disruptions, being flooded, power outages, suffocation, returning home, and being homeless; emergency management topics include rescue, measures, fire fighting, mutual assistance, emergency, evacuation, soldiers, flood control, and flood prevention; disaster-bearing carrier topics include passengers, carriages, people, airports, tunnels, subway entrances, cities, the entire line, buses, roads, citizens, communities, and trains; positive evaluation topics include hope, peace, cheer up, stay safe, hang in there, pay attention to safety, and blessings; negative evaluation topics include terrible, heartbreaking, distressing, and doomed.

[0133] In this application embodiment, clothing includes clothes, quilts, clothing, cotton quilts, bedding, tents, mattresses, blankets, and folding beds; food includes food, diet, snacks, instant noodles, mineral water, food, drinking water, boxed meals, crispy noodles, and bread; housing includes dwellings, tents, houses, homes, homelessness, resettlement, refuge, and flooding; transportation includes roads, traffic, highways, bridges, obstruction, blockage, collapse, flooding, traffic jams, accidents, congestion, disputes, and service disruptions.

[0134] Step 502: Based on the feature vectors, the preprocessed target online public opinion data is divided into the corresponding topic category and livelihood category.

[0135] In this embodiment of the application, public opinion data is divided into corresponding topic categories and livelihood categories.

[0136] Step 503: Count the number of public opinions for each topic category and livelihood category in each time period, and plot the time-quantity change curve.

[0137] In this embodiment of the application, the number of public opinions for each topic category and the livelihood category within each time period is counted.

[0138] Step 504: Based on the plotted time-quantity change curve, obtain the key concerns and needs of the public in each time period.

[0139] In this embodiment of the application, the changes in public attention and public needs are obtained by plotting the time-quantity change curve, which improves the pertinence of the subway public opinion management department when issuing documents.

[0140] In terms of emotion, topic, and people's livelihood analysis, it helps to identify negative emotions among the public and passengers and take timely measures to guide public opinion.

[0141] Step 104: Based on statistical data, develop the best emergency rescue plan.

[0142] In this embodiment of the application, by statistically analyzing the amount of public opinion information related to the above four aspects of needs, the emergency assistance most needed by the public at each time period can be obtained, thereby helping the subway emergency management department to take timely emergency measures and ensure the public's basic needs for food, clothing, housing and transportation.

[0143] This application embodiment utilizes web crawling technology to specify the keywords for crawling online public opinion regarding subway emergencies, identify key information such as the release time, location, and text content in online public opinion information, and analyze online public opinion regarding subway emergencies from the perspectives of time, space, sentiment, and hot topics. This allows for the identification of the emergency assistance most needed by the public at different time periods, thereby helping subway emergency management departments to take timely emergency measures, ensuring the public's basic needs for food, clothing, housing, and transportation, and providing support for the safe operation of the subway system, public opinion analysis, and public opinion guidance.

[0144] Figure 6 This is a block diagram of a subway emergency network public opinion analysis device 600 based on Weibo text, as illustrated in an exemplary embodiment of this application. It includes an acquisition module 610, a preprocessing module 620, a multi-class analysis module 630, and an output module 640.

[0145] The acquisition module 610 is used to formulate online public opinion collection rules and determine the target online public opinion data in subway emergencies based on the online public opinion collection rules. The online public opinion collection rules include collection time rules, collection keyword rules and collection field rules.

[0146] The preprocessing module 620 is used to preprocess the target online public opinion data and extract the feature vectors of the text within the target online public opinion data;

[0147] The multi-analysis module 630 is used to perform time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis on the preprocessed target online public opinion data, and to statistically analyze the data of each analysis to obtain statistical data.

[0148] Output module 640 is used to develop the best emergency rescue plan based on statistical data.

[0149] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0150] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0151] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for analyzing online public opinion regarding subway emergencies based on microblog text, characterized in that, include: Formulate rules for collecting online public opinion, and determine the target online public opinion data for subway emergencies based on the rules. The rules for collecting online public opinion include rules for collection time, rules for collection keywords, and rules for collection fields. The target online public opinion data is preprocessed to extract feature vectors from the text within the target online public opinion data. This includes: performing Chinese word segmentation on the target online public opinion data to determine a vocabulary list; traversing each word in the vocabulary list, removing a word if its text length is less than 2 or it is in a stop word list; and calculating the TF-IDF weights of the remaining words after removal. The steps for calculating the TF-IDF weights of any word are as follows: ， in, The number of times the word appears in this piece of public opinion text. M represents the number of words remaining in the public opinion text after removing stop words, and M represents the number of entries in the public opinion text. The number of public opinion texts containing this term; feature weights are constructed based on the TF-IDF weights and the remaining terms, and are formulated as follows: ； For a target network public opinion data, the remaining words after the removal process are sorted from largest to smallest according to the TF-IDF weights, and the words and TF-IDF weights of the first Vim position are retained to complete the dimensionality reduction of the feature vector. The preprocessed target online public opinion data is subjected to time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis, and statistical analysis is performed on the data from each analysis to obtain statistical data. The process of sentiment analysis on the preprocessed target online public opinion data includes: calculating the sentiment value of the target online public opinion data according to the Naive Bayes algorithm; classifying the sentiment value of the target online public opinion data according to preset rules to determine the sentiment tendency, wherein the sentiment tendency includes negative sentiment, neutral sentiment, and positive sentiment; classifying the preprocessed target online public opinion data according to preset time intervals and calculating the sentiment tendency of each time group after the division, wherein if the sentiment tendency of a certain time group is negative or the difference between the sentiment value of a certain time group and the sentiment value of the previous time group is not less than a preset value, the subway public opinion management department releases an official Weibo post; determining the sentiment value of the target online public opinion data based on geographical location information, wherein if the sentiment value of a station or section is lower than the preset value, the reasons for the sentiment value of the station or section are analyzed, and emergency rescue measures are taken for the station or section. Based on the statistical data, develop the best emergency response plan.

2. The method according to claim 1, characterized in that, The data collection time rules include: Determine the analysis requirements for the aforementioned subway emergencies; If the analysis requirement is to analyze historical emergencies, the data collection time range is selected from one week before the occurrence of the historical emergencies to one week after the end of the historical emergencies. If the analysis requirement is real-time monitoring, the online public opinion information is collected in real time, and the collection method is to collect once every Δt minutes, with each collection containing the online public opinion information from the previous Δt minutes.

3. The method according to claim 1, characterized in that, The keyword collection rules include: The number of mandatory words and the number of arbitrarily included words in online public opinion information text are determined, wherein the number of mandatory words is greater than 0, and the number of arbitrarily included words is not limited.

4. The method according to claim 1, characterized in that, The rules for the collected fields include: According to the rules of the social media platform, target fields of online public opinion information in the subway emergency are extracted. The target fields include the poster's nickname, the poster's authentication information, the posting time, the posting content, the number of reposts, the number of comments, and the number of likes.

5. The method according to claim 1, characterized in that, The time series analysis of the preprocessed target online public opinion data includes: Based on the preprocessed target online public opinion data, determine the number of online public opinion events in each time period and plot the time-quantity change curve; By analyzing the time-quantity change curve, the likelihood of an online public opinion outbreak can be determined.

6. The method according to claim 1, characterized in that, The spatial analysis of the preprocessed target online public opinion data includes: Obtain the geographic location information of the preprocessed target network public opinion data, wherein the geographic location information includes subway line information, subway station information and subway section information; By combining the text content of the preprocessed target network public opinion data, the geographical location information is identified to determine distress messages and refuge information; The number of distress messages and evacuation messages in the geographical location information is counted, and the number and location of the information are plotted on the subway network map using a heat map method.

7. The method according to claim 1, characterized in that, The analysis of topics and public sentiment on the preprocessed target online public opinion data includes: Based on the public safety triangle theory model, the topic is divided into disaster environment, disaster impact, emergency management, disaster-bearing carriers, positive evaluation and negative evaluation, and people's livelihood is divided into clothing, food, housing and transportation. Based on the feature vectors, the preprocessed target online public opinion data is divided into corresponding topic categories and livelihood categories; The number of public opinions for each of the aforementioned topic categories and livelihood categories within each time period was statistically analyzed, and a time-quantity change curve was plotted. Based on the plotted time-quantity change curve, the public's focus and needs for each period can be obtained.

8. A device for analyzing online public opinion regarding subway emergencies based on microblog text, characterized in that: include: The acquisition module is used to formulate online public opinion collection rules and determine the target online public opinion data in the subway emergency based on the online public opinion collection rules. The online public opinion collection rules include collection time rules, collection keyword rules and collection field rules. The preprocessing module is used to preprocess the target online public opinion data and extract feature vectors of the text within the target online public opinion data. This includes: performing Chinese word segmentation on the target online public opinion data to determine a vocabulary list; traversing each word in the vocabulary list, removing a word if its text length is less than 2 or it is in a stop word list; and calculating the TF-IDF weights of the remaining words after removal. The steps for calculating the TF-IDF weights of any word are as follows: ， in, The number of times the word appears in this piece of public opinion text. M represents the number of words remaining in the public opinion text after removing stop words, and M represents the number of entries in the public opinion text. The number of public opinion texts containing this term; feature weights are constructed based on the TF-IDF weights and the remaining terms, and are formulated as follows: ； For a target network public opinion data, the remaining words after the removal process are sorted from largest to smallest according to the TF-IDF weights, and the words and TF-IDF weights of the first Vim position are retained to complete the dimensionality reduction of the feature vector. The system includes multiple analysis modules for performing time series analysis, spatial analysis, sentiment analysis, and topic and livelihood analysis on the preprocessed target online public opinion data, and for statistical analysis of each analysis data to obtain statistical data. The sentiment analysis process for the preprocessed target online public opinion data includes: calculating the sentiment value of the target online public opinion data using the Naive Bayes algorithm; classifying the sentiment value of the target online public opinion data according to preset rules to determine the sentiment tendency, where the sentiment tendency includes negative sentiment, neutral sentiment, and positive sentiment; classifying the preprocessed target online public opinion data according to preset time intervals and calculating the sentiment tendency of each time group after classification, where if the sentiment tendency of a certain time group is negative or the difference between the sentiment value of a certain time group and the sentiment value of the previous time group is not less than a preset value, the subway public opinion management department will release an official Weibo post; determining the sentiment value of the target online public opinion data based on geographical location information, where if the sentiment value of a station or section is lower than the preset value, the reasons for the sentiment value of the station or section are analyzed, and emergency rescue measures are taken for the station or section. The output module is used to formulate the best emergency rescue plan based on the statistical data.

Images