A method and system for analyzing appeals of passenger normalization problems of urban rail transit, and a storage medium

Abstract

This invention relates to a method for analyzing routine passenger complaints in urban rail transit, comprising: S1, multi-source data collection; S2, passenger complaint analysis; and S3, early warning mechanism construction and automatic departmental push notifications. This invention also relates to a system for analyzing routine passenger complaints in urban rail transit, comprising a data collection module, a passenger complaint analysis module, and an early warning module. This invention addresses the problem of insufficient data dimensions in existing systems, significantly improving the comprehensiveness of analysis results; based on large language models and deep semantic analysis technology, it can accurately identify the true content and sentiment of passenger complaints, achieving high-precision classification and key sentence extraction; it improves the efficiency of operating units in handling complaints; and it presents overall complaint trends through visual charts, providing real-time decision-making basis for operation management.

Description

Technical Field

[0001] This invention belongs to the field of transportation service technology, and relates to traffic data collection and analysis, specifically to a method, system and storage medium for analyzing the routine problems and demands of urban rail transit passengers. Background Technology

[0002] With the acceleration of urbanization, rail transit, as a crucial component of urban public transportation, bears increasing passenger load. Currently, passenger complaints and feedback primarily rely on data manually recorded in customer service hotlines and operations management systems. This traditional method suffers from numerous problems, including incomplete data collection and low processing efficiency. Meanwhile, with the widespread use of social media (such as Weibo and Zhihu), more and more passengers tend to express their travel experiences and complaints online, leading to a diversification and fragmentation of data sources. However, how to efficiently and accurately acquire, integrate, and analyze multi-source data from social media platforms, customer service recordings, and passenger surveys has become a pressing technical challenge for rail transit management departments.

[0003] Currently, the passenger demand analysis system closest to this invention is mainly based on the following two technical frameworks: 1. Rule-based passenger complaint handling system: This system relies on a predefined set of rules to classify and statistically analyze passenger complaints and feedback. Its advantages are that it is simple to implement, easy to deploy, and suitable for rapid processing in normal scenarios. However, it has problems such as poor rule scalability and insufficient flexibility, making it difficult to deal with sudden or complex passenger complaints. At the same time, the cost of rule updates and maintenance is high, and it cannot fully reflect the dynamic changes of multi-source data.

[0004] 2. Passenger Request Analysis Model Based on Machine Learning: This type of system uses historical complaint data to train a machine learning model to automatically classify and perform sentiment analysis on passenger feedback. Its advantage lies in its ability to make predictions driven by data and achieve relatively fast responses in some scenarios; however, this method heavily relies on the comprehensiveness and accuracy of the data, and the results generated by the model often lack sufficient interpretability and transparency, making it difficult to gain the trust of managers and achieve effective verification in practical applications.

[0005] In summary, existing technologies suffer from the following main problems: ① Lack of comprehensiveness: Existing systems typically employ a single rule set or machine learning model, failing to fully integrate the advantages of multi-source data and making it difficult to comprehensively understand passenger demands; ② Low efficiency in data integration and processing: Due to the diverse and dispersed nature of data sources, traditional methods are inefficient in the data collection, cleaning, and fusion processes, easily leading to the omission of key information; ③ Inadequate early warning mechanisms: When faced with complex and dynamically changing passenger feedback, existing systems struggle to identify high-risk demands in a timely and accurate manner, resulting in delayed problem warnings and impacting operational management effectiveness. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, system and storage medium for analyzing the routine problems and demands of urban rail transit passengers. By comprehensively utilizing distributed data acquisition, advanced semantic understanding, knowledge graph construction and time series analysis technologies, it can achieve efficient integration and in-depth analysis of passenger feedback information, thereby providing more accurate and real-time decision support for rail transit operation management.

[0007] The technical problem solved by this invention is achieved through the following technical solution: A method for analyzing the routine issues and demands of urban rail transit passengers, the steps of which are as follows: S1. Multi-source data collection: Complete the collection of data from social media platforms, customer service hotline recordings, and passenger survey data to form a multi-source dataset and transcribe the above multi-source dataset into a structured dataset, including serial number, timestamp, passenger ID, passenger request text content, sentiment tendency classification, and request keywords. Among them, the sentiment tendency classification includes strong positive crisis, mild positive crisis, neutral crisis, mild negative crisis, and strong negative crisis. S2. Passenger Request Analysis: Based on semantic understanding and sentiment analysis technology using a large language model, the text content of passenger requests is analyzed to determine the sentiment tendency classification and request keywords; S2.1 Text semantic vectorization and sentiment classification; This involves vectorizing the passenger request text content of each record in the structured dataset. Input the encoder of the fine-tuned large language model to obtain its high-dimensional semantic vector representation: ; in: Encoding functions for large language models, For the corresponding semantic vector; semantic vectors Input the sentiment classification header, and calculate the probability distribution of each sentiment category through a fully connected layer and a Softmax function: ; in: and For the weights and biases of the sentiment classification layer, These are probability vectors for five types of emotions; The emotional tendency classification of this passenger's request is determined based on the principle of maximum probability: ; in, Indicates the first The passenger's appeal text belongs to the emotion category. Predicted probability, set of sentiment categories The obtained sentiment categories are written into the corresponding records in the structured dataset. Field; S2.2, Keyword extraction and semantic point annotation; obtaining semantic vectors Based on this, the importance of each word in the text is scored using the token-level output and attention weights of the large language model. Combined with a keyword dictionary and named entity recognition results from the urban rail transit field, words related to fault phenomena, service issues, safety hazards, and passenger experience are filtered. By setting weight thresholds and deduplication rules, a keyword set is formed for each text. in: This is a keyword extraction function. This represents the number of keywords selected in the text. The keyword set is serialized and written into the corresponding record in the structured dataset. This completes the determination of the sentiment and annotation of the key points of the passenger's request text. S3, Early Warning Mechanism Construction and Automatic Departmental Push; Sentiment Tendency Classification Output in S2 With keyword set Based on this, risk assessments are conducted on passenger requests and the requests are then forwarded to the relevant departments. S3.1 Calculation of early warning level; Construction of emotional risk scoring function Keyword risk weighting function The emotional risk scoring function assigns different base risk values ​​based on the degree of negativity of the emotional category. For example, strong negative / crisis is mapped to the highest risk score, and strong positive is mapped to the lowest risk score. The keyword risk weight function calculates the risk weight based on the category of the keyword (such as safety, equipment failure, service experience) and its frequency and location in the text. Combining these two factors yields a comprehensive warning score for each passenger request. ; in: and An adjustable weighting coefficient is used to balance the impact of emotional intensity and problem category on overall risk; based on the warning score. Based on preset grading thresholds, passenger requests are divided into several warning levels; S3.2 Department Mapping and Automatic Push; Based on Keyword Sets The business themes included automatically map passenger requests to different functional departments of the urban rail transit operation organization; Keywords related to train operation, train intervals, and delay information are mapped to the operations dispatch department; Keywords related to equipment such as signals, vehicles, platform facilities, air conditioning, and lighting are mapped to the Equipment Maintenance Department; Keywords related to ticket purchase, inquiry, signage, and service attitude are mapped to the passenger service department; Map security-related keywords such as public order, conflict, and suspicious persons to the security department; Keywords related to the overall riding experience and satisfaction ratings are mapped to the Passenger Experience Department; Keywords related to route operation plans and capacity adjustment suggestions are mapped to the route planning department; After determining the warning level and target department, the system will access the corresponding passenger request text records. Sentimental tendency classification Key words for appeal timestamp and warning level These messages are combined into warning messages and automatically pushed to the target department's business system or work terminal via message queue or API call, enabling real-time warnings and cross-departmental collaborative processing of passengers' routine requests.

[0008] Moreover, S1 specifically refers to: S1.1 Social Media Platform Data Collection: Using pre-configured API calls or web crawlers, automatically capture publicly available text content related to urban rail transit from Weibo, WeChat official account comment sections, app review sections, and forums within set time intervals. Record the capture time, source platform, and user identification information to obtain the raw social media data set. The original data is deduplicated, ad-free, and spam-free filtered. Meaningless symbols are removed and emojis are uniformly encoded. Fields with strong privacy attributes are desensitized. Then, sentence segmentation and word segmentation algorithms are used to split the long text into several independently analyzable segments to obtain a cleaned social media text set. S1.2 Customer Service Hotline Recording and Transcription: Voice call recordings generated by the customer service center of urban rail transit operating companies are used as voice data sources. These recordings are indexed and stored according to call record number, caller ID, and call start and end times, forming a voice data set. After preprocessing the speech data with noise suppression, echo cancellation and volume normalization, the speech recognition model is called to transcribe the speech content into a text sequence and retain the timestamp information corresponding to the speech frame; the transcribed text is then processed with punctuation completion, colloquial expression standardization and misspelling correction, and interjections and stop words are filtered to obtain a set of hotline transcribed texts. S1.3 Passenger Survey and Evaluation Data Collection: Collect passenger evaluation information on train punctuality, comfort, service attitude, and sense of security through offline questionnaires, online satisfaction surveys, and station self-service evaluation terminals. Convert paper questionnaires into electronic text through OCR or manual input. Perform format verification and field completion on the electronic survey data, delete invalid questionnaires and extreme outliers, and obtain a structured survey data set. S1.4 Unified Encoding and Structured Storage of Multi-Source Data: Align the social media text, hotline transcripts, and survey data obtained in S1.1–1.3 according to unified field specifications, and establish a structure including serial numbers. timestamp Passenger identification Passenger Request Text Content Sentimental tendency classification and keywords of appeal Multi-source structured datasets: ; Among them: In the data construction phase, fields and This is the state to be identified; set it to null for subsequent analysis and filling.

[0009] A system for analyzing the routine issues and demands of urban rail transit passengers, comprising: The data acquisition module is used to collect social media data, customer service hotline recordings, and passenger survey data. The passenger feedback analysis module is used to classify and analyze passenger feedback data based on large language models and knowledge graphs. The early warning module is used to trigger early warnings based on the frequency and intensity of complaints, and push them to the relevant responsible parties.

[0010] A computer-readable storage medium stores a routine problem complaint analysis program for urban rail transit passengers, wherein when the routine problem complaint analysis program for urban rail transit passengers is executed by a processor, the program implements the steps of the routine problem complaint analysis method for urban rail transit passengers.

[0011] The advantages and beneficial effects of this invention are as follows: 1. Strong integration of multi-source data: Covering three major information sources, including social media, hotline recordings, and questionnaires, it solves the problem of insufficient data dimensions in existing systems and significantly improves the comprehensiveness of analysis results.

[0012] 2. Intelligent analysis and early warning capabilities: Based on large language models and deep semantic analysis technology, it can accurately identify passengers' true demands and emotional tendencies, and achieve high-precision classification and key sentence extraction.

[0013] 3. Intelligent push and decision support: Automatically match the responsible department and push the results to improve the efficiency of the operation unit in handling requests; present the overall request trend through visual charts to provide real-time decision-making basis for operation management. Attached Figure Description

[0014] Figure 1 This is a flowchart of the social media information crawler of the present invention; Figure 2 This is a flowchart of the multi-source data acquisition process of the present invention; Figure 3 This is a flowchart of the passenger demand analysis and push process of the present invention; Figure 4 This is a diagram illustrating how a web crawler retrieves information from social media pages. Figure 5 Example image of the results retrieved for passenger requests; Figure 6 A graph showing the results of the sentiment analysis of passenger demands; Figure 7 A statistical analysis chart of passenger sentiment; Figure 8 A screenshot showing the push notifications for passenger feedback analysis results. Detailed Implementation

[0015] The present invention will be further described in detail below through specific embodiments. The following embodiments are merely descriptive and not limiting, and should not be used to limit the scope of protection of the present invention.

[0016] An innovative method for analyzing the routine issues and demands of urban rail transit passengers is described, with the following steps: A method for analyzing the routine issues and demands of urban rail transit passengers, the steps of which are as follows: S1. Multi-source data collection: Complete the collection of data from social media platforms, customer service hotline recordings, and passenger survey data to form a multi-source dataset and transcribe the above multi-source dataset into a structured dataset, including serial number, timestamp, passenger ID, passenger request text content, sentiment tendency classification, and request keywords. Among them, the sentiment tendency classification includes strong positive crisis, mild positive crisis, neutral crisis, mild negative crisis, and strong negative crisis. S1.1 Social Media Platform Data Collection: Using pre-configured API calls or web crawlers, automatically capture publicly available text content related to urban rail transit from Weibo, WeChat official account comment sections, app review sections, and forums within set time intervals. Record the capture time, source platform, and user identification information to obtain the raw social media data set. The original data is deduplicated, ad-free, and spam-free filtered. Meaningless symbols are removed and emojis are uniformly encoded. Fields with strong privacy attributes are desensitized. Then, sentence segmentation and word segmentation algorithms are used to split the long text into several independently analyzable segments to obtain a cleaned social media text set. S1.2 Customer Service Hotline Recording and Transcription: Voice call recordings generated by the customer service center of urban rail transit operating companies are used as voice data sources. These recordings are indexed and stored according to call record number, caller ID, and call start and end times, forming a voice data set. After preprocessing the speech data with noise suppression, echo cancellation and volume normalization, the speech recognition model is called to transcribe the speech content into a text sequence and retain the timestamp information corresponding to the speech frame; the transcribed text is then processed with punctuation completion, colloquial expression standardization and misspelling correction, and interjections and stop words are filtered to obtain a set of hotline transcribed texts. S1.3 Passenger Survey and Evaluation Data Collection: Collect passenger evaluation information on train punctuality, comfort, service attitude, and sense of security through offline questionnaires, online satisfaction surveys, and station self-service evaluation terminals. Convert paper questionnaires into electronic text through OCR or manual input. Perform format verification and field completion on the electronic survey data, delete invalid questionnaires and extreme outliers, and obtain a structured survey data set. S1.4 Unified Encoding and Structured Storage of Multi-Source Data: Align the social media text, hotline transcripts, and survey data obtained in S1.1–1.3 according to unified field specifications, and establish a structure including serial numbers. timestamp Passenger identification Passenger Request Text Content Sentimental tendency classification and keywords of appeal Multi-source structured datasets: ; Among them: In the data construction phase, fields and This is the state to be identified; set it to null for subsequent analysis and filling.

[0017] S2. Passenger Request Analysis: Based on semantic understanding and sentiment analysis technology using a large language model, the text content of passenger requests is analyzed to determine the sentiment tendency classification and request keywords; S2.1 Text semantic vectorization and sentiment classification; This involves vectorizing the passenger request text content of each record in the structured dataset. Input the encoder of the fine-tuned large language model to obtain its high-dimensional semantic vector representation: in: Encoding functions for large language models, For the corresponding semantic vector; semantic vectors Input the sentiment classification header, and calculate the probability distribution of each sentiment category through a fully connected layer and a Softmax function: ; in: and For the weights and biases of the sentiment classification layer, These are probability vectors for five types of emotions; The emotional tendency classification of this passenger's request is determined based on the principle of maximum probability: ; in, Indicates the first The passenger's appeal text belongs to the emotion category. Predicted probability, set of sentiment categories The obtained sentiment categories are written into the corresponding records in the structured dataset. Field; S2.2, Keyword extraction and semantic point annotation; obtaining semantic vectors Based on this, the importance of each word in the text is scored using the token-level output and attention weights of the large language model. Combined with a keyword dictionary and named entity recognition results from the urban rail transit field, words related to fault phenomena, service issues, safety hazards, and passenger experience are filtered. By setting weight thresholds and deduplication rules, a keyword set is formed for each text. in: This is a keyword extraction function. This represents the number of keywords selected in the text. The keyword set is serialized and written into the corresponding record in the structured dataset. This completes the determination of the sentiment and annotation of the key points of the passenger's request text. S3, Early Warning Mechanism Construction and Automatic Departmental Push; Sentiment Tendency Classification Output in S2 With keyword set Based on this, risk assessments are conducted on passenger requests and the requests are then forwarded to the relevant departments. S3.1 Calculation of early warning level; Construction of emotional risk scoring function Keyword risk weighting function The emotional risk scoring function assigns different base risk values ​​based on the degree of negativity of the emotional category. For example, strong negative / crisis is mapped to the highest risk score, and strong positive is mapped to the lowest risk score. The keyword risk weight function calculates the risk weight based on the category of the keyword (such as safety, equipment failure, service experience) and its frequency and location in the text. Combining these two factors yields a comprehensive warning score for each passenger request. ; in: and An adjustable weighting coefficient is used to balance the impact of emotional intensity and problem category on overall risk; based on the warning score. Based on preset grading thresholds, passenger requests are divided into several warning levels; S3.2 Department Mapping and Automatic Push; Based on Keyword Sets The business themes included automatically map passenger requests to different functional departments of the urban rail transit operation organization; Keywords related to train operation, train intervals, and delay information are mapped to the operations dispatch department; Keywords related to equipment such as signals, vehicles, platform facilities, air conditioning, and lighting are mapped to the Equipment Maintenance Department; Keywords related to ticket purchase, inquiry, signage, and service attitude are mapped to the passenger service department; Map security-related keywords such as public order, conflict, and suspicious persons to the security department; Keywords related to the overall riding experience and satisfaction ratings are mapped to the Passenger Experience Department; Keywords related to route operation plans and capacity adjustment suggestions are mapped to the route planning department; After determining the warning level and target department, the system will access the corresponding passenger request text records. Sentimental tendency classification Key words for appeal timestamp and warning level These messages are combined into warning messages and automatically pushed to the target department's business system or work terminal via message queue or API call, enabling real-time warnings and cross-departmental collaborative processing of passengers' routine requests.

[0018] Moreover, S1 specifically refers to: A system for analyzing the routine issues and demands of urban rail transit passengers, comprising: The data acquisition module is used to collect social media data, customer service hotline recordings, and passenger survey data. The passenger feedback analysis module is used to classify and analyze passenger feedback data based on large language models and knowledge graphs. The early warning module is used to trigger early warnings based on the frequency and intensity of complaints, and push them to the relevant responsible parties.

[0019] A computer-readable storage medium stores a routine problem complaint analysis program for urban rail transit passengers, wherein when the routine problem complaint analysis program for urban rail transit passengers is executed by a processor, the program implements the steps of the routine problem complaint analysis method for urban rail transit passengers.

[0020] In this embodiment, the data acquisition module is used to automate the collection of multi-source data. This module includes a social media interface, a hotline recording processing unit, and a passenger survey questionnaire management system. The social media interface connects to Weibo, WeChat official accounts, and Xiaohongshu platforms via API to capture passenger comments, complaints, and suggestions related to rail transit in real time; the hotline recording processing unit uses speech recognition technology to transcribe customer service call recordings into text information; and the questionnaire management system is responsible for collecting raw data from online and offline passenger satisfaction surveys to form structured feedback information. The entire data acquisition process is as follows: Figure 1 and Figure 2 As shown, the interface for a web crawler is as follows: Figure 4 As shown, the data obtained after setting the crawler parameters is uniformly stored in the database to provide data support for subsequent analysis. The crawling results are as follows. Figure 5 As shown.

[0021] In this invention, the passenger feedback analysis module is responsible for intelligent classification and in-depth analysis of the raw feedback data. This module performs semantic understanding of the collected text data based on a large language model, and combines natural language processing technology to perform word segmentation, syntactic analysis, and sentiment analysis on passenger feedback. The implementation process is as follows: Figure 3 As shown. Through a dedicated database for the rail transit sector, the system can identify the business type of the request, such as operation scheduling, facility maintenance, safety management, and service attitude, and further automatically match the responsible department. The sentiment analysis submodule can also determine the emotional intensity of each feedback, providing a basis for subsequent early warning mechanisms. Figure 6 and Figure 7 The results shown are the sentiment analysis results and statistical results of passenger demands.

[0022] The early warning module dynamically identifies potential problem trends based on passenger complaint analysis and promptly pushes relevant information to the corresponding responsible parties. This module constructs a tiered early warning system, setting complaint frequency thresholds and emotional intensity thresholds based on time windows. When a certain type of complaint surges in a short period or is accompanied by strong negative emotions, the system will trigger an early warning at the corresponding level. Simultaneously, by combining the mapping rules between passenger complaints and responsible departments, the system can achieve a precise push mechanism, automatically distributing the corresponding complaint content and analysis conclusions to relevant departments (such as facilities management, customer service, and safety supervision) for rapid response. Early warning information supports graphical display, allowing managers to monitor changes in complaint trends through a visual interface, aiding decision-making and resource allocation. Figure 8 The image shown is the passenger complaint analysis results push page. The system automatically counts the number of complaints pending, in process, and processed, and intuitively displays the recent complaint processing process and the historical complaint processing status of each department.

[0023] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for analyzing appeals of urban rail transit passenger normalization problems, characterized in that: The steps of the method are as follows: S1. Multi-source data collection: Complete the collection of data from social media platforms, customer service hotline recordings, and passenger survey data to form a multi-source dataset and transcribe the above multi-source dataset into a structured dataset, including serial number, timestamp, passenger ID, passenger request text content, sentiment tendency classification, and request keywords. Among them, the sentiment tendency classification includes strong positive crisis, mild positive crisis, neutral crisis, mild negative crisis, and strong negative crisis. S2. Passenger Request Analysis: Based on semantic understanding and sentiment analysis technology using a large language model, the text content of passenger requests is analyzed to determine the sentiment tendency classification and request keywords; S2.1 Text semantic vectorization and sentiment classification; This involves vectorizing the passenger request text content of each record in the structured dataset. Input the encoder of the fine-tuned large language model to obtain its high-dimensional semantic vector representation: in: Encoding functions for large language models, For the corresponding semantic vector; semantic vectors Input the sentiment classification header, and calculate the probability distribution of each sentiment category through a fully connected layer and a Softmax function: ； in: and For the weights and biases of the sentiment classification layer, These are probability vectors for five types of emotions; The emotional tendency classification of this passenger's request is determined based on the principle of maximum probability: ； in, Indicates the first The passenger's appeal text belongs to the emotion category. Predicted probability, set of sentiment categories The obtained sentiment categories are written into the corresponding records in the structured dataset. Field; S2.2, Keyword extraction and semantic point annotation; obtaining semantic vectors Based on this, the importance of each word in the text is scored using the token-level output and attention weights of the large language model. Combined with a keyword dictionary and named entity recognition results from the urban rail transit field, words related to fault phenomena, service issues, safety hazards, and passenger experience are filtered. By setting weight thresholds and deduplication rules, a keyword set is formed for each text. in: This is a keyword extraction function. This represents the number of keywords selected in the text. The keyword set is serialized and written into the corresponding record in the structured dataset. This completes the determination of the sentiment and annotation of the key points of the passenger's request text. S3, Early Warning Mechanism Construction and Automatic Departmental Push; Sentiment Tendency Classification Output in S2 With keyword set Based on this, risk assessments are conducted on passenger requests and the requests are then forwarded to the relevant departments. S3.1 Calculation of early warning level; Construction of emotional risk scoring function Keyword risk weighting function The emotional risk scoring function assigns different base risk values ​​based on the degree of negativity of the emotional category. For example, strong negative / crisis is mapped to the highest risk score, and strong positive is mapped to the lowest risk score. The keyword risk weight function calculates the risk weight based on the category of the keyword (such as safety, equipment failure, service experience) and its frequency and location in the text. Combining these two factors yields a comprehensive warning score for each passenger request. ； in: and An adjustable weighting coefficient is used to balance the impact of emotional intensity and problem category on overall risk; based on the warning score. Based on preset grading thresholds, passenger requests are divided into several warning levels; S3.2 Department Mapping and Automatic Push; Based on Keyword Sets The business themes included automatically map passenger requests to different functional departments of the urban rail transit operation organization; Keywords related to train operation, train intervals, and delay information are mapped to the operations dispatch department; Keywords related to equipment such as signals, vehicles, platform facilities, air conditioning, and lighting are mapped to the Equipment Maintenance Department; Keywords related to ticket purchase, inquiry, signage, and service attitude are mapped to the passenger service department; Map security-related keywords such as public order, conflict, and suspicious persons to the security department; Keywords related to the overall riding experience and satisfaction ratings are mapped to the Passenger Experience Department; Keywords related to route operation plans and capacity adjustment suggestions are mapped to the route planning department; After determining the warning level and target department, the system will access the corresponding passenger request text records. Sentimental tendency classification Key words for appeal timestamp and warning level These messages are combined into warning messages and automatically pushed to the target department's business system or work terminal via message queue or API call, enabling real-time warnings and cross-departmental collaborative processing of passengers' routine requests.

2. The method for analyzing the routine issues and demands of urban rail transit passengers according to claim 1, characterized in that: Specifically, S1 is: S1.1 Social Media Platform Data Collection: Using pre-configured API calls or web crawlers, automatically capture publicly available text content related to urban rail transit from Weibo, WeChat official account comment sections, app review sections, and forums within set time intervals. Record the capture time, source platform, and user identification information to obtain the raw social media data set. The original data is deduplicated, ad-free, and spam-free filtered. Meaningless symbols are removed and emojis are uniformly encoded. Fields with strong privacy attributes are desensitized. Then, sentence segmentation and word segmentation algorithms are used to split the long text into several independently analyzable segments to obtain a cleaned social media text set. S1.2 Customer service hotline recording and speech transcription; Voice call recordings generated by the customer service center of urban rail transit operating companies are used as voice data sources. They are indexed and stored according to call record number, caller ID, and call start and end time to form a voice data set. After preprocessing the speech data with noise suppression, echo cancellation and volume normalization, the speech recognition model is called to transcribe the speech content into a text sequence and retain the timestamp information corresponding to the speech frame; the transcribed text is then processed with punctuation completion, colloquial expression standardization and misspelling correction, and interjections and stop words are filtered to obtain a set of hotline transcribed texts. S1.3 Passenger Survey and Evaluation Data Collection: Collect passenger evaluation information on train punctuality, comfort, service attitude, and sense of security through offline questionnaires, online satisfaction surveys, and station self-service evaluation terminals. Convert paper questionnaires into electronic text through OCR or manual input. Perform format verification and field completion on the electronic survey data, delete invalid questionnaires and extreme outliers, and obtain a structured survey data set. S1.4 Unified Encoding and Structured Storage of Multi-Source Data: Align the social media text, hotline transcripts, and survey data obtained in S1.1–1.3 according to unified field specifications, and establish a structure including serial numbers. timestamp Passenger identification Passenger Request Text Content Sentimental tendency classification and keywords of appeal Multi-source structured datasets: ； Among them: In the data construction phase, fields and This is the state to be identified; set it to null for subsequent analysis and filling.

3. A system for analyzing the routine issues and demands of urban rail transit passengers, comprising: The data acquisition module is used to collect social media data, customer service hotline recordings, and passenger survey data. The passenger feedback analysis module is used to classify and analyze passenger feedback data based on large language models and knowledge graphs. The early warning module is used to trigger early warnings based on the frequency and intensity of complaints, and push them to the relevant responsible parties.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a routine problem-solving program for urban rail transit passengers. When the routine problem-solving program for urban rail transit passengers is executed by a processor, it implements the steps of the routine problem-solving method for urban rail transit passengers as described in any one of claims 1 and 2.

Images