Public complaint automatic processing method and system in social governance

Abstract

The invention is suitable for the technical field of computer processing, and particularly relates to a public complaint automatic processing method and system in social governance, and the method comprises the steps: obtaining target complaint data and a complaint data package; performing feature extraction on the target complaint data to obtain target features; determining a complaint change feature based on the target feature and the complaint data packet; determining associated features based on the target complaint data and complaint template data; wherein the complaint template data comprises typical complaint types and modes summarized in social governance, and the correlation features are used for representing correlation between the target complaint data and the complaint template data; performing feature fusion on the target feature, the complaint change feature and the associated feature to obtain a comprehensive feature; and performing priority classification processing on the comprehensive features to obtain the priority of the target complaint. The method effectively solves the problem that the processing efficiency is low due to the fact that complaint data is large and difficult to screen quickly.

Description

Technical Field

[0001] The present application belongs to the field of computer processing technology, and in particular, relates to a method and system for automatically handling public complaints in social governance. Background Art

[0002] Public complaints play an important role in social governance. They are not only a channel for the public to express their demands, but also a key way for the government and social managers to obtain public opinions and improve services. Through public complaints, the public can reflect the various problems they encounter in life, work, environment, etc., so that the governance can understand the needs and problems of the society more clearly.

[0003] In the existing technology, public complaints in social governance often involve huge amounts of data, and the needs and problems of different complainants may vary greatly. Manual or simple automated methods make it difficult to screen and process a large number of complaints within a limited time, thus delaying the response to urgent or important complaints.

[0004] To sum up, when handling public complaints in social governance, there is a problem of low processing efficiency due to the large amount of complaint data and the difficulty in rapid screening. Summary of the invention

[0005] The embodiments of the present application provide a method and system for automatically handling public complaints in social governance, which can solve the problem in related technologies of low processing efficiency due to the difficulty in quickly screening the large amount of complaint data when handling public complaints in social governance.

[0006] In a first aspect, the embodiment of the present application provides a method for automatically handling public complaints in social governance, including: Obtain target complaint data and complaint data packages; wherein the target complaint data is used to represent the public complaint data of a certain complainant in social governance, and the complaint data package is used to represent all historical public complaint data in social governance; Extracting features from the target complaint data to obtain target features; Determining complaint change characteristics based on the target characteristics and the complaint data packet; Based on the target complaint data and the complaint template data, determining correlation features; wherein the complaint template data includes typical complaint types and patterns summarized in social governance, and the correlation features are used to characterize the correlation between the target complaint data and the complaint template data; Performing feature fusion on the target feature, the complaint change feature and the associated feature to obtain a comprehensive feature; The comprehensive features are prioritized and classified to obtain the priority of the target complaint.

[0007] The above technical solutions in the embodiments of the present application have at least the following technical effects: The present application provides a method for automatically processing public complaints in social governance. First, the target complaint data and complaint data packets are obtained, and then the target complaint data are feature extracted to obtain the target features. Based on the target features and complaint data packets, the complaint change features are determined, and based on the target complaint data and complaint template data, the associated features are determined, which is conducive to knowing the correlation between the target complaint data and the complaint template data. Then, the target features, complaint change features and associated features are feature fused to obtain comprehensive features. Finally, the comprehensive features are prioritized and processed to obtain the priority of the target complaint. This method reduces the manual screening of a large amount of complaint data through automated feature extraction and priority classification, and reduces the workload and the risk of human error. This method can reasonably sort the urgency and importance of complaints, which helps the government or social governance departments to give priority to high-priority issues when resources are limited, and avoid delaying the resolution of key issues. This method realizes efficient processing and intelligent management of large-scale complaint data, effectively solves the problem of low processing efficiency due to the difficulty in rapid screening of large amounts of complaint data, and provides strong support for social governance.

[0008] In a possible implementation manner of the first aspect, extracting features from the target complaint data to obtain target features includes: extracting text features from the target complaint data; Calculate the sentiment score and sentiment category of the target complaint data using a sentiment analysis model to obtain sentiment features; The text feature and the sentiment feature are integrated to obtain the target feature.

[0009] In a possible implementation manner of the first aspect, determining the complaint change feature based on the target feature and the complaint data packet includes: Determining the complaint type of the target complaint according to the text features in the target features; Extracting historical complaint data of the same complaint type as the target complaint from the complaint data packet; According to the timestamps of the historical complaint data, the historical complaint data are arranged in ascending order of time to obtain a complaint time series; Based on the complaint time series, time domain features and frequency domain features are extracted, and the time domain features and the frequency domain features are integrated to obtain the complaint change features.

[0010] In a possible implementation manner of the first aspect, extracting time domain features and frequency domain features based on the complaint time series includes: Based on the complaint time series, the number of complaints in each window is calculated using a sliding window; based on the number of complaints in each window, the complaint change rate of each window is calculated; based on the change rate of each window, the time domain feature is extracted; Perform Fourier transform on the complaint time series to obtain frequency domain features.

[0011] In a possible implementation manner of the first aspect, determining the associated features based on the target complaint data and the complaint template data includes: Calculate the similarity between the target complaint data and each complaint template in the complaint template data using a similarity algorithm to obtain a similarity score for each complaint template; Selecting the complaint template with the highest similarity score from the similarity scores of each complaint template; Arrange the similarity scores of each complaint template from high to low to obtain a similarity score sequence, and extract the similarity scores of at least the first two complaint templates from the similarity score sequence; Based on the similarity score of each complaint template, calculate the variance of the similarity score and the mean of the similarity score; The similarity score of the complaint template with the highest similarity score, the similarity scores of at least the first two complaint templates in the similarity score sequence, the variance of the similarity scores and the mean of the similarity scores are integrated to obtain the associated feature.

[0012] In a possible implementation of the first aspect, the priority classification processing of the comprehensive features to obtain the priority of the target complaint includes: Based on each sub-decision tree in the decision tree model, the comprehensive features are prioritized and classified to obtain sub-results of each sub-decision tree; The sub-results of each sub-decision tree are scored to obtain the score value of each category result; wherein the category result is used to represent the priority category among the sub-results of all sub-decision trees; Based on the score values ​​of each category result, the priority of the target complaint is determined.

[0013] In a possible implementation manner of the first aspect, the scoring process of the sub-results of each sub-decision tree to obtain a score value of each category result includes: Classify the sub-results of each sub-decision tree to obtain at least two category results; Calculating the number of sub-results of each of the at least two category results, and determining the number of sub-results of each of the at least two category results as a score value of each category result; The step of determining the priority of the target complaint based on the score value of each sub-decision tree includes: The category result with the highest score value is determined as the priority of the target complaint.

[0014] In a possible implementation manner of the first aspect, the scoring process of the sub-results of each sub-decision tree to obtain a score value of each category result includes: Obtaining a category score value; wherein the category score value is used to represent the score value corresponding to each category result; Selecting a score value corresponding to each category result of the at least two category results from the category score values; Calculate a score value for each category result based on the score value corresponding to each category result of the at least two category results and the number of sub-results of each category result of the at least two category results; The step of determining the priority of the target complaint based on the score value of each sub-decision tree includes: Calculate the total score of the at least two category results based on the score value of each category result in the at least two category results; Calculate an average of the at least two category results based on the total score of the at least two category results; Comparing the average of the at least two category results with the scoring threshold to obtain a comparison result; According to the comparison result, the priority of the target complaint is determined.

[0015] In a possible implementation manner of the first aspect, the method further includes: Get a sample complaint dataset; Priority marking is performed on each sample complaint data in the sample complaint data set to obtain a priority label for each sample complaint data; The sample complaint data set and the priority label of each sample complaint data are divided into a training set and a test set, and the decision tree model is trained based on the training set and the test set.

[0016] In a second aspect, the embodiment of the present application provides a system for automatically handling public complaints in social governance, including: An acquisition unit, used to acquire target complaint data and complaint data packets; wherein the target complaint data is used to represent the public complaint data of a certain complainant in social governance, and the complaint data packets are used to represent all historical public complaint data in social governance; A target feature obtaining unit, used for extracting features from the target complaint data to obtain target features; A complaint change feature determination unit, configured to determine a complaint change feature based on the target feature and the complaint data packet; A correlation feature determination unit, configured to determine a correlation feature based on the target complaint data and the complaint template data; wherein the correlation feature is used to characterize the correlation between the target complaint data and the complaint template data; A comprehensive feature obtaining unit, used for performing feature fusion on the target feature, the complaint change feature and the associated feature to obtain a comprehensive feature; The priority determination unit is used to perform priority classification processing on the comprehensive features to obtain the priority of the target complaint.

[0017] In the third aspect, an embodiment of the present application provides an automatic processing device for public complaints in social governance, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of the embodiments of the first aspect when executing the computer program.

[0018] It can be understood that the beneficial effects of the second to third aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 It is a flowchart of a method for automatically handling public complaints in social governance provided by an embodiment of the present application; Figure 2 It is a schematic diagram of the implementation flow of step S200 and step 300 in the method for automatically handling public complaints in social governance provided in an embodiment of the present application; Figure 3 It is a schematic diagram of the implementation process of step S600 in the method for automatically handling public complaints in social governance provided in an embodiment of the present application; Figure 4 This is another implementation flow diagram of step S600 in the method for automatically handling public complaints in social governance provided in an embodiment of the present application; Figure 5 It is a schematic diagram of the structure of the automatic public complaint processing system in social governance provided by the embodiment of the present application; Figure 6 It is a structural diagram of the automatic public complaint processing equipment in social governance provided in the embodiment of the present application. DETAILED DESCRIPTION

[0021] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0022] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0023] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0024] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0025] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0026] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0027] In related technologies, public complaints in social governance often involve huge amounts of data, and the needs and problems of different complainants may vary greatly. Manual or simple automated methods are difficult to screen and process a large number of complaints within a limited time, thus delaying the response to urgent or important complaints. To solve the above problems, the embodiment of the present application provides a method and system for automatically processing public complaints in social governance. In this method, the target complaint data and complaint data packets are first obtained, and then the target complaint data is feature extracted to obtain the target features. Based on the target features and complaint data packets, the complaint change features are determined, and based on the target complaint data and complaint template data, the associated features are determined, which is conducive to knowing the correlation between the target complaint data and the complaint template data, and then the target features, complaint change features and associated features are feature fused to obtain comprehensive features, and finally the comprehensive features are prioritized and processed to obtain the priority of the target complaint. This method reduces the manual screening of a large amount of complaint data through automated feature extraction and priority classification, and reduces the workload and the risk of human error. This method can reasonably sort the urgency and importance of complaints, which helps the government or social governance department to give priority to high-priority issues when resources are limited, and avoid delaying the resolution of key issues. This method realizes efficient processing and intelligent management of large-scale complaint data, effectively solves the problem of low processing efficiency due to the difficulty in rapid screening of large amounts of complaint data, and provides strong support for social governance.

[0028] The method for automatic handling of public complaints in social governance provided in the embodiment of the present application can be applied to the automatic handling device for public complaints in social governance. At this time, the automatic handling device for public complaints in social governance is the executor of the method for automatic handling of public complaints in social governance provided in the embodiment of the present application. The embodiment of the present application does not impose any restrictions on the specific type of the automatic handling device for public complaints in social governance.

[0029] For example, the automatic public complaint processing equipment in social governance can be a mobile phone, a tablet computer, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), a desktop computer, a smart screen, a computer, a laptop computer, a handheld computing device, a customer premise equipment (CPE) and / or other equipment used for communicating on wireless systems and next-generation communication systems, such as mobile terminals in 5G networks or mobile terminals in future evolved public land mobile networks (PLMN).

[0030] In order to better understand the method for automatic handling of public complaints in social governance provided by the embodiment of the present application, the specific implementation process of the method for automatic handling of public complaints in social governance provided by the embodiment of the present application is exemplarily introduced below.

[0031] Figure 1 A schematic flow chart of a method for automatically handling public complaints in social governance provided by an embodiment of the present application is shown. The method for automatically handling public complaints in social governance includes: S100, obtaining target complaint data and complaint data package, wherein the target complaint data is used to represent the public complaint data of a certain complainant in social governance, and the complaint data package is used to represent all historical public complaint data in social governance.

[0032] It can be understood that the target complaint data is used to characterize the public complaints submitted by a specific complainant (or complaint group) in the process of social governance, and may include the complainant's basic information (such as identity, region, time period, etc.), complaint content, etc. The complaint data package represents all historical public complaint data in social governance, and contains complaint records of multiple complainants. The characteristics of the complaint data package are wide coverage and large data volume, and may include information on all complaints within a time range for the system to conduct overall analysis and statistics.

[0033] For example, the complainant's complaint records within a specific time range can be extracted from the complaint management system or complaint database based on the complainant's unique identifier (such as the complainant's ID, mobile phone number, ID number, etc.) through an API interface or database query. The extracted fields may include complaint type, complaint content, submission time, etc.

[0034] All public complaint data can be filtered out from the complaint database or data warehouse according to time, region, complaint type and other conditions to obtain a complaint data package.

[0035] The complaint database can choose a relational database (such as MySQL, PostgreSQL) or a NoSQL database (such as MongoDB, Elasticsearch). For large-scale complaint data, distributed storage technologies such as Hadoop, Spark, etc. can be used to manage and process data.

[0036] Through this step, the target complaint data and complaint data packages can be accurately obtained, providing strong support for complaint management, problem troubleshooting and decision-making in social governance.

[0037] S200, extracting features from target complaint data to obtain target features.

[0038] It can be understood that the target features may include basic features, sentiment features, complaint theme features, etc. The basic features may include time features, such as complaint time (specific to year / month / day / hour), complaint frequency (the number of complaints submitted by complainants within a fixed time window), time interval (the time interval between consecutive complaints); spatial features, such as complaint location (specific address or normalized regional information), regional distribution (whether the complaints of complainants are concentrated in certain specific regions).

[0039] Exemplarily, for time features, timestamps can be extracted from the target complaint data, that is, the complaint time; the number of complaints in the target complaint data can be counted to calculate the complaint frequency; the time interval between consecutive complaints can be calculated based on the timestamps. For spatial features, location keywords in the target complaint data, such as complaint location, regional distribution, etc., can be extracted using word segmentation and keyword extraction algorithms (such as TF-IDF, TextRank). Sentiment features can be extracted from the target complaint data using sentiment analysis tools (such as VADER, BERT). Themes can be extracted from the target complaint data using topic models (such as LDA, Transformer models).

[0040] Through this step, structured and actionable information can be refined from the target complaint data, and the target features lay a foundation for subsequent classification, priority division, and complaint management optimization.

[0041] In a possible implementation, please refer to Figure 2 , S200, perform feature extraction on the target complaint data to obtain target features, including: S210, extract text features from the target complaint data.

[0042] Exemplarily, the target complaint data can be preprocessed: cleaning the data, stop words (such as "de", "le", "shi") can be removed, and irrelevant content such as special characters, punctuation marks, and spaces can be removed; word segmentation can be performed using word segmentation tools (such as Jieba, NLTK) on the complaint text to obtain a list of words.

[0043] The text features may include keyword feature vectors. TF-IDF (Term Frequency - Inverse Document Frequency) can be used to extract important keywords in the text to obtain TF-IDF weights, that is, keyword feature vectors; topic feature vectors. LDA (Latent Dirichlet Allocation) can be used to extract the text topic distribution to generate topic feature vectors. For example, the correlation degree of a certain text with topic 1 is 0.6, and the correlation degree with topic 2 is 0.4; text feature vectors. Word embedding models (such as Word2Vec, BERT) can be used to convert the text into vectors to capture semantic relationships.

[0044] S220, using the sentiment analysis model to calculate the sentiment score and sentiment category of the target complaint data to obtain the sentiment features.

[0045] Exemplarily, a pre-trained sentiment analysis model (such as TextBlob, VADER, BERT sentiment classification model) or a custom classification model (a binary / multi-classification model trained based on historical complaint data) can be used. The target complaint data is input into the sentiment analysis model, and the sentiment analysis model can calculate the sentiment score of the target complaint data. The sentiment score range is [-1, 1]. A value greater than zero indicates positive sentiment, a value less than zero indicates negative sentiment, and a value equal to zero indicates neutral sentiment. Based on the sentiment score, the text in the target complaint data can be classified into specific emotions (such as anger, anxiety, satisfaction, etc.). The sentiment intensity score can be calculated by the model to measure the intensity of the complainant's emotional expression. For example, the sentiment score is -0.75, the sentiment category is anger (negative), and the sentiment intensity is 0.8.

[0046] S230, integrating text features and sentiment features to obtain target features.

[0047] For example, the text features and sentiment features can be merged by column to form a unified feature vector, i.e., the target feature. , emotional characteristics (The emotion category is anger, which is coded as 1), then the target feature after feature concatenation is .

[0048] The weighted fusion method can be used to fuse text features and sentiment features to obtain the target features, that is, , is the weight of the text feature, is the weight of the sentiment feature, and Both can be set manually.

[0049] S300, determining complaint change characteristics based on target characteristics and complaint data packets.

[0050] It can be understood that complaint change characteristics may include time dimension change characteristics, which compare the target characteristics with the time trend in the complaint data package; spatial dimension change characteristics, which compare the target complaint location distribution with the regional distribution of the overall data package; content dimension change characteristics, which compare the target complaint content with the historical complaint data content in the complaint data package.

[0051] Exemplarily, the time dimension change characteristics may include trend characteristics, which can calculate the change curve of the complaint frequency in the target feature over time and the overall complaint frequency curve in the complaint data packet, and use regression analysis or time series models (such as ARIMA) to extract trend information; periodic characteristics, which can use periodic analysis (such as FFT) to detect periodic changes in target complaints.

[0052] Spatial dimension change characteristics: Based on the complaint locations in the target features, a regional heat map can be generated, and similarity analysis (such as JS divergence and cosine similarity) can be performed between the regional heat map and the overall heat map of the complaint data package. Clustering algorithms such as DBSCAN can be used to detect the migration or diffusion of target complaint hotspots.

[0053] Features of content dimension changes may include changes in keyword frequency. Keyword extraction (such as TF-IDF, Word2Vec) is performed on historical complaint data in the complaint data packet, and the keywords in the target feature are compared with the keywords in the complaint data packet to analyze the trend of new or disappearing keywords. For topic changes, models such as LDA and BERT can be used to extract the topic distribution in the complaint data packet, calculate the KL divergence or JS divergence between the topic distribution in the target feature and the topic distribution in the complaint data packet, and analyze the topic shift.

[0054] Through this step, we can comprehensively analyze the dynamic changes of target characteristics and complaint data packets to derive the complaint change characteristics, providing strong support for trend prediction, problem diagnosis and resource optimization in social governance.

[0055] In one possible implementation, see Figure 2 , S300, based on the target features and the complaint data packet, determine the complaint change features, including: S310, determining the complaint type of the target complaint according to the text features in the target features.

[0056] For example, keyword matching can be used to determine the complaint type of the target complaint. Use a keyword table or classification rules to match the text features of the target complaint (such as keywords or topic distribution) with predefined complaint types. For example, if the matching rule contains keywords such as noise and nuisance, the complaint type is an environmental problem; if it contains keywords such as parking and congestion, the complaint type is a traffic problem.

[0057] The classification model can be used to determine the complaint type of the target complaint. The text feature vector is input into a trained text classification model (such as BERT, Random Forest), and the text classification model outputs the complaint type label.

[0058] S320, extracting historical complaint data of the same complaint type as the target complaint from the complaint data packet.

[0059] For example, according to the determined target complaint type, all historical complaint records with the same target complaint type can be filtered from the complaint data package. For example, if the complaint type is environmental issues, all records with the complaint type of environmental issues in the complaint data package are extracted. The extracted fields can include complaint ID, timestamp (submission time), location or region (optional).

[0060] S330, according to the timestamps of the historical complaint data, the historical complaint data are arranged in ascending order of time to obtain a complaint time series.

[0061] For example, the historical complaint data can be sorted in ascending order by the timestamp field to construct a time series ,in Represents a time point of a complaint. The time series can be discretized (such as counting by day or hour) to construct a time-complaint volume pair. ,in Indicated in The number of complaints during the time period.

[0062] S340, based on the complaint time series, extract time domain features and frequency domain features, and integrate the time domain features and the frequency domain features to obtain complaint change features.

[0063] It can be understood that the time domain features describe the statistical characteristics of the time series over time, and the frequency domain features describe the periodicity and frequency characteristics of the time series.

[0064] For example, the time-complaint volume Extract the maximum and minimum values ​​of the number of complaints to reflect the peak and valley values ​​of complaints. , calculate the average complaint frequency ,Right now, Based on the time-complaint volume and average complaint frequency , calculate the standard deviation of complaint frequency ,Right now, Linear regression can be used to fit the time-complaint curve and extract the trend coefficient, that is, ,in Represents the trend coefficient. The maximum number of complaints, the minimum number of complaints, the average complaint frequency, the standard deviation of the complaint frequency, and the trend coefficient can be combined to obtain the time domain characteristics.

[0065] The time-complaint pair can be transformed using Fourier transform (FFT) Convert from time domain to frequency domain to get the spectrum amplitude , represents the amplitude of the frequency component. Select the frequency corresponding to the maximum amplitude from the spectrum amplitude , that is, the main frequency, and according to the main frequency , calculate the period corresponding to the main frequency, that is, Based on the spectrum amplitude , calculate the total spectrum energy The main frequency, the period corresponding to the main frequency and the sum of the spectrum energy are combined to obtain the frequency domain features.

[0066] The time domain features and frequency domain features can be combined to form a complete complaint change feature.

[0067] By integrating time domain and frequency domain features, the complaint change characteristics obtained can accurately describe the dynamics of complaint behavior and provide high-value analytical support for social governance.

[0068] Optionally, S340, based on the complaint time series, extracting time domain features and frequency domain features, including: S341, based on the complaint time series, use a sliding window to calculate the number of complaints in each window; calculate the complaint change rate of each window based on the number of complaints in each window; extract time domain features based on the change rate of each window.

[0069] It can be understood that a sliding window is a fixed-length window that moves on a time series and is used to extract statistical features of a local time period. The parameters of a time window include the window size, the time span within the window (such as 1 hour, 1 day); and the sliding step, the time interval between each window movement (such as 1 hour).

[0070] For example, a sliding window can be applied to the time series, and the number of complaints can be calculated for each window, that is, ,in, Display Window The number of complaints Indicates the window size. The sliding window outputs a time series of the number of complaints, indicating the total number of complaints in each window.

[0071] It can be understood that the complaint change rate of the sliding window reflects the changing trend and magnitude of the number of complaints between windows.

[0072] For example, the complaint change rate of each window is determined by the difference between the number of complaints in the previous window and the number of complaints in the current window, and the calculation formula is: ,in, Display Window The change rate of complaints Display Window The number of complaints in the previous window, if , then the complaint change rate can be set to 0 or a defined constant. After the complaint change rate of each window is calculated, the change rate sequence of the sliding window can be obtained. .

[0073] It can be understood that the time domain features are used to describe the change rate series Statistical properties and trends over time.

[0074] For example, the sliding window can be used to change the rate sequence , calculate the mean of the complaint change rate ,Right now, ; can be based on the average of the complaint change rate , calculate the standard deviation of the complaint change rate ,Right now, ; The rate of change sequence of the sliding window can be Extract the maximum complaint change rate and minimum complaint change rate ; The range of change can be calculated, that is, the maximum complaint change rate Minus the minimum complaint change rate , get the range of change; the change rate sequence can be Linear regression is performed to obtain the trend coefficient, which represents the rate of change of complaints. The mean of the rate of change of complaints can be , standard deviation of complaint change rate , Maximum complaint change rate , minimum complaint change rate , variation range and trend coefficient are combined to obtain the time domain characteristics.

[0075] S342, performing Fourier transform on the complaint time series to obtain frequency domain features.

[0076] For example, the complaint number sequence of the sliding window can be represented as a spectrum by fast Fourier transform (FFT), and the spectrum amplitude can be obtained: , represents the amplitude of the frequency component. Select the frequency with the largest spectrum amplitude from the spectrum amplitude (dominant frequency), indicating the dominant periodicity of the number of complaints. , calculate the period corresponding to the main frequency, that is, Based on the spectrum amplitude , calculate the total spectrum energy The main frequency, the period corresponding to the main frequency and the sum of the spectrum energy are combined to obtain the frequency domain features.

[0077] S400, determining correlation features based on the target complaint data and the complaint template data, wherein the complaint template data includes typical complaint types and patterns summarized in social governance, and the correlation features are used to characterize the correlation between the target complaint data and the complaint template data.

[0078] It can be understood that complaint template data can be historical data or expert experience summary, and can include typical complaint types, such as noise disturbance, traffic congestion, public facility failure, etc.; pattern characteristics, summarized complaint type characteristics, such as common keywords, typical complaint behaviors (frequency, time distribution); processing strategy, summarized best processing plan or result distribution.

[0079] The associated features may include content relevance, which is the similarity between the target complaint content and the typical complaint types in the template; and pattern relevance, which is the similarity between the behavior pattern (such as frequency and time distribution) of the target complaint and the pattern summarized in the template.

[0080] Exemplarily, content relevance may include: keyword matching, which can extract keywords from the target complaint data (using TF-IDF or word frequency analysis) and calculate the matching rate of the target complaint and template complaint keywords; text similarity, which can use text vectorization models (such as Word2Vec, TF-IDF vectors) to represent the target complaint data and template complaint data, and calculate the cosine similarity between the two; topic distribution similarity, which can use topic models such as LDA to extract the topic distribution of the target complaint data and template data, and calculate the similarity of the topic distribution (KL divergence, JS divergence).

[0081] Pattern relevance may include: time distribution similarity, which can extract the time distribution of target complaints (such as high-frequency complaint time periods) and calculate the intersection of the time patterns of target complaints and template summaries; frequency distribution similarity, which can compare the frequency of target complaints with the complaint frequency distribution summarized in the template, and use KL divergence or cosine similarity to measure the distribution difference between the two; spatial distribution similarity, which can map the geographic information of target complaints to the hotspot area distribution of the template, and calculate whether the target complaints are concentrated in the hotspot areas defined by the template.

[0082] A weighted fusion method or a feature splicing method can be used to fuse content relevance and pattern relevance into a relevance feature.

[0083] Through this step, the correlation features between the target complaint data and the complaint template data can be efficiently extracted, providing support for further classification processing, pattern matching and priority analysis.

[0084] In a possible implementation, S400, based on the target complaint data and the complaint template data, determining the associated features includes: S410, using a similarity algorithm to calculate the similarity between the target complaint data and each complaint template in the complaint template data, and obtain a similarity score for each complaint template.

[0085] For example, when both the target complaint data and the complaint template data are text, the text embedding method and the cosine similarity method can be used to calculate the similarity between the target complaint data and each complaint template in the complaint template data. The word embedding model (such as TF-IDF, Word2Vec, BERT) can be used to convert each complaint template in the target complaint data and complaint template data into a vector, and then the cosine similarity between the vector corresponding to the target complaint data and the vector corresponding to each template can be calculated.

[0086] When both the target complaint data and the complaint template data are text and similarity needs to be calculated based on word frequency, you can use TF-IDF to vectorize the target complaint text and each template text, and then calculate the cosine similarity between the target complaint text and each template text as the similarity score.

[0087] When the text of the target complaint data and complaint template data is complex and involves deep semantics, you can use a pre-trained model such as BERT to generate embedding vectors for the target complaint data and each template data, and then calculate the cosine similarity between the embedding vectors.

[0088] Traverse the template data set, calculate the similarity between the target complaint data and each template using the above method, and obtain the similarity score and similarity score sequence of each complaint template ,in Indicates the target complaint data and The similarity score of the complaint templates.

[0089] S420, selecting a complaint template with the highest similarity score from the similarity scores of each complaint template.

[0090] For example, the similarity score sequence can be Filter out the maximum similarity score .

[0091] S430, arranging the similarity scores of each complaint template from high to low to obtain a similarity score sequence, and extracting the similarity scores of at least the first two complaint templates from the similarity score sequence.

[0092] For example, the complaint templates can be sorted from high to low according to their similarity scores to obtain a sorted similarity score sequence, that is, , .from Extract at least the first two similarity scores from .

[0093] S440, based on the similarity score of each complaint template, calculate the variance of the similarity score and the mean of the similarity score.

[0094] For example, the similarity score sequence can be , calculate the mean of the similarity scores ,Right now, ; According to the mean of similarity scores , calculate the variance of the similarity score ,Right now, .

[0095] S450, integrating the similarity score of the complaint template with the highest similarity score, the similarity scores of at least the first two complaint templates in the similarity score sequence, the variance of the similarity scores and the mean of the similarity scores to obtain a correlation feature.

[0096] For example, the maximum similarity score , at least the first two similarity scores , the mean of the similarity scores and the variance of the similarity score Merge to get the associated features, that is, .

[0097] Through the above steps, the target complaint data can be efficiently correlated with the complaint template data for analysis, and comprehensive correlation features can be generated to provide support for further decision-making.

[0098] S500, feature fusion is performed on the target feature, complaint change feature and related feature to obtain a comprehensive feature.

[0099] For example, since the value ranges of different feature dimensions may vary greatly, the target features, complaint change features, and associated features can be standardized. Min-Max Scaling can be used to scale numerical features to the range of [0, 1]; Z-score standardization can be used to normalize the mean of unevenly distributed features.

[0100] The feature concatenation method can be used to concatenate the target features, complaint change features, and associated features by column to form a comprehensive feature vector, that is, , Represents comprehensive characteristics, represents the target feature, Indicates the characteristics of complaint changes, Represents an associated feature.

[0101] The feature weighting method can be used to manually set the weights corresponding to the target feature, complaint change feature, and associated feature, and then perform weighted summation of the target feature, complaint change feature, and associated feature according to their corresponding weights to obtain the comprehensive feature, that is, , is the weight of the target feature, is the weight of the complaint change feature, is the weight of the associated feature.

[0102] Through this step, the target features, complaint change features and associated features can be integrated to generate a comprehensive integrated feature vector, providing strong support for subsequent complaint management decisions.

[0103] S600, performing priority classification processing on the comprehensive features to obtain the priority of the target complaint.

[0104] Exemplarily, the rule engine classification method can be used to prioritize the comprehensive features. Complaints are prioritized using business rules. For example, the rule design can be high priority, where the complaint type is public safety or a serious social impact issue, the complaint content keywords match emergency, safety, and major, and the time trend change rate is greater than the preset threshold; medium priority, where the complaint involves regional issues and the change rate is moderate; low priority, where the complaint is highly repetitive, has a high similarity to the template, and has a fixed solution. Rules can be written using logical conditions to directly output priority categories based on the value of the comprehensive features.

[0105] Machine learning classification methods can be used to prioritize the comprehensive features. Appropriate machine learning models can be selected, such as decision trees, random forests, gradient boosting trees (XGBoost / LightGBM), etc. Sample complaint data is labeled with complaint priorities, and the labeled sample complaint data is used as a training set to train the selected machine learning model to obtain a classification model. The comprehensive feature vector is used as the input feature of the classification model, and the classification model can predict the priority category of the target complaint.

[0106] Through this method, efficient and accurate target complaint priority classification and processing can be achieved, providing strong support for social governance and resource allocation.

[0107] In a possible implementation, S600 performs priority classification processing on the comprehensive features to obtain the priority of the target complaint, including: S610, based on each sub-decision tree in the decision tree model, priority classification processing is performed on the comprehensive features to obtain a sub-result of each sub-decision tree.

[0108] It can be understood that a decision tree is a hierarchical classification model in which each node can divide the data according to a certain condition of the comprehensive features and finally reach the leaf node. In this case, each child decision tree can independently perform priority classification processing and output a classification result.

[0109] Exemplarily, the comprehensive feature vector can be input into each sub-decision tree, each sub-decision tree independently processes the comprehensive features of the target complaint, and each sub-decision tree outputs a classification result (sub-result). For example, the classification result of decision tree 1 is high priority, the classification result of decision tree 2 is medium priority, and the classification result of decision tree 3 is high priority.

[0110] S620, scoring the sub-results of each sub-decision tree to obtain a score value for each category result, wherein the category result is used to represent the priority category among the sub-results of all sub-decision trees.

[0111] Exemplarily, the sub-results of each sub-decision tree may be mapped to numerical scores for subsequent priority determination. Scoring rules for sub-results may be preset, and sub-results of all sub-decision trees may be aggregated according to the scoring rules for sub-results, and scores may be calculated by category.

[0112] Assume that the scoring rule for sub-results is that the score corresponding to high priority is 3 points, the score corresponding to medium priority is 2 points, and the score corresponding to low priority is 1 point. The classification result of decision tree 1 is high priority (3 points), the classification result of decision tree 2 is medium priority (2 points), and the classification result of decision tree 3 is high priority (3 points). Then the total score of high priority is 6 points, the total score of medium priority is 2 points, and the total score of low priority is 0 points.

[0113] S630, determining the priority of the target complaint based on the score value of each category result.

[0114] Exemplarily, the score values ​​of each category are sorted, and the category with the highest score is selected as the final priority of the target complaint. For example, if the total score of high priority is 6 points, the total score of medium priority is 2 points, and the total score of low priority is 0 points, then the priority of the target complaint is high priority.

[0115] If the scores are the same (for example, both high priority and medium priority have a score of 5), the average score of the sub-decision trees of the corresponding category can be calculated; higher weights can be given to certain decision trees based on business rules; and the probability of the target category can be calculated based on the distribution of sub-results.

[0116] Through the above steps, classification fusion can be performed on the results of multiple sub-decision trees to obtain accurate priority decisions to meet the needs of multiple scenarios.

[0117] Optionally, see Figure 3 , S620, scoring the sub-results of each sub-decision tree to obtain the score value of each category result, including: S621, classify the sub-results of each sub-decision tree to obtain at least two category results.

[0118] Exemplarily, the sub-results of all sub-decision trees can be classified according to priority categories (such as high priority, medium priority, and low priority) to obtain a set of each priority category. For example, sub-decision tree 1 is a high priority, sub-decision tree 2 is a medium priority, sub-decision tree 3 is a high priority, and sub-decision tree 4 is a low priority, then the high priority includes sub-decision tree 1 and sub-decision tree 3, the medium priority includes sub-decision tree 2, and the low priority includes sub-decision tree 4.

[0119] S622: Calculate the number of sub-results of each of the at least two category results, and determine the number of sub-results of each of the at least two category results as a score value of each category result.

[0120] Exemplarily, the number of sub-results in each priority category can be counted and the number of sub-results can be used as the score value of the category. For example, high priority includes sub-decision tree 1 and sub-decision tree 3, medium priority includes sub-decision tree 2, and low priority includes sub-decision tree 4, then the score value of high priority is 2, the score value of medium priority is 1, and the score value of low priority is 1.

[0121] Optionally, see Figure 3 , S630, based on the score value of each sub-decision tree, determine the priority of the target complaint, including: S631, determine the category result with the highest score value as the priority of the target complaint.

[0122] For example, the scores of each priority category can be compared, and the category with the highest score can be selected as the priority of the target complaint. If the scores are the same, some categories can be given higher priority (such as high priority first) according to business rules; one can be randomly selected from the category with the highest score. For example, the score of high priority is 2, the score of medium priority is 1, and the score of low priority is 1. The high priority score is the highest, so the priority of the target complaint is high priority.

[0123] Optionally, see Figure 4 , S620, scoring the sub-results of each sub-decision tree to obtain the score value of each category result, including: S6201, obtaining a category score value, wherein the category score value is used to represent the score value corresponding to each category result.

[0124] It can be understood that the category score value can be a fixed score for each category (such as high priority, medium priority, low priority) to measure the importance of the category.

[0125] Exemplarily, the category score value can be manually set and input through the human-computer interaction window. For example, the high priority score value is 3 points, the medium priority score value is 2 points, and the low priority score value is 1 point.

[0126] S6202: Select a score value corresponding to each category result of at least two category results from the category score values.

[0127] For example, assume that the category score is: 3 points for high priority, 2 points for medium priority, and 1 point for low priority. If the category results are high priority and medium priority, then the score corresponding to high priority is 3 points, and the score corresponding to medium priority is 2 points.

[0128] S6203: Calculate a score value of each category result based on the score value corresponding to each category result in the at least two category results and the number of sub-results of each category result in the at least two category results.

[0129] For example, the score value of each category can be calculated according to the score value of each category result and the number of sub-results of each category result. The calculation formula is: ,in represents the score value of the category result, Indicates the score value of the category result, Indicates the number of sub-results in a category result. For example, the category result includes high priority (score value = 3, sub-results number = 4), medium priority (score value = 2, number = 3). The score value of high priority is , the medium priority score is .

[0130] Optionally, see Figure 4 , S630, based on the score value of each sub-decision tree, determine the priority of the target complaint, including: S6301: Calculate a total score of at least two category results based on the score of each category result in at least two category results.

[0131] Exemplarily, the score values ​​of at least two category results are summed to obtain a total score value. For example, the score value of high priority is 12, and the score value of medium priority is 6, then the total score value is 18.

[0132] S6302: Calculate an average value of at least two category results based on the total scores of at least two category results.

[0133] Exemplarily, based on the total score, the average value of at least two category results is calculated. For example, the category results are high priority and low priority, and the total score is 18, then the average value is 18 / 2=9.

[0134] S6303: Compare the average value of the at least two category results with the scoring threshold to obtain a comparison result.

[0135] Exemplarily, a scoring threshold range can be preset, and the average value can be compared with the scoring threshold range. For example, the scoring threshold range can be 5 to 7, and when the average value is greater than the scoring threshold range, the priority is high priority; when the average value is less than the scoring threshold range, the priority is low priority; when the average value is within the scoring threshold range, the priority is medium priority.

[0136] S6304: Determine the priority of the target complaint based on the comparison result.

[0137] Exemplarily, assuming that the average value is 9, the score threshold range is 5 to 7, and the average value is greater than the score threshold range, then the priority of the target complaint is high priority.

[0138] In a possible implementation, the method for automatically handling public complaints in social governance also includes: S10, obtain a sample complaint dataset.

[0139] Exemplarily, historical complaint data may be obtained from a complaint database and used as sample complaint data.

[0140] S20, marking the priority of each sample complaint data in the sample complaint data set to obtain a priority label for each sample complaint data.

[0141] For example, a priority label can be added to each sample complaint data to indicate the urgency or importance of the complaint. Each piece of data can be prioritized according to business rules or expert opinions. For example, high priority can be related to public safety and major social impact; medium priority can be general service complaints or ordinary problems; low priority can be small impact and more personal problems. Staff can mark the priority of each piece of data; data can be automatically marked by the rule engine. For example, if the complaint text is that the construction noise in the community is loud at night, affecting rest, it can be marked as high priority; if the complaint text is that the property service attitude is poor and the problem has not been solved, it can be marked as medium priority.

[0142] S30, dividing the sample complaint data set and the priority label of each sample complaint data into a training set and a test set, and training the decision tree model based on the training set and the test set.

[0143] For example, the data set can be divided into a training set and a test set in a certain ratio (such as 80% training set and 20% test set), and a random seed or stratified sampling method can be used to ensure that the distribution of each priority category in the training set and the test set is as balanced as possible.

[0144] You can choose a suitable algorithm, such as CART (classification and regression tree), ID3 or C4.5 decision tree algorithms. Convert the comprehensive features in the sample complaint data into a numerical format suitable for decision tree model processing, and use the converted training set data and the priority labels corresponding to the training set data to train the decision tree model. The decision tree starts from the root node, selects an optimal feature each time, and divides the sample into different branches until the leaf node. The decision tree is divided based on maximizing information gain (or other metrics, such as the Gini index) and generating a tree structure according to the splitting effect of the feature. After the model training is completed, use the test set to verify the accuracy, recall rate and F1 score of the model.

[0145] Through the above steps, the decision tree model can be efficiently trained and tested, thereby achieving accurate prediction of priority classification.

[0146] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0147] Corresponding to the method for automatic handling of public complaints in social governance described in the above embodiments, the embodiments of the present application also provide a system for automatic handling of public complaints in social governance, and each unit of the system can implement each step of the method for automatic handling of public complaints in social governance. Figure 5 A structural block diagram of the automatic public complaint handling system in social governance provided by an embodiment of the present application is shown. For the sake of ease of explanation, only the parts related to the embodiment of the present application are shown.

[0148] Reference Figure 5 , the system comprises: The acquisition unit is used to acquire target complaint data and complaint data packages. The target complaint data is used to represent the public complaint data of a complainant in social governance, and the complaint data package is used to represent all historical public complaint data in social governance.

[0149] The target feature obtaining unit is used to extract features from the target complaint data to obtain target features.

[0150] The complaint change feature determination unit is used to determine the complaint change feature based on the target feature and the complaint data packet.

[0151] The correlation feature determination unit is used to determine the correlation feature based on the target complaint data and the complaint template data, wherein the correlation feature is used to characterize the correlation between the target complaint data and the complaint template data.

[0152] The comprehensive feature acquisition unit is used to fuse the target features, complaint change features and related features to obtain comprehensive features.

[0153] The priority determination unit is used to perform priority classification processing on the comprehensive features to obtain the priority of the target complaint.

[0154] It should be noted that the information interaction, execution process and other contents between the above-mentioned units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0155] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0156] The present application also provides a device for automatically processing public complaints in social governance. Figure 6 This is a schematic diagram of the structure of an automatic public complaint processing device in social governance provided by an embodiment of the present application. Figure 6 As shown, the public complaint automatic processing device 6 in social governance of this embodiment includes: at least one processor 60 ( Figure 6 Only one is shown), at least one memory 61 ( Figure 6 Only one is shown in the figure) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the automatic public complaint processing device 6 in social governance implements the steps in any of the above-mentioned embodiments of the automatic public complaint processing method in social governance, or implements the functions of each unit in the above-mentioned device embodiments.

[0157] Exemplarily, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 62 in the public complaint automatic processing device 6 in social governance.

[0158] The automatic public complaint processing device 6 in social governance can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The automatic public complaint processing device in social governance can include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art can understand that Figure 6 It is only an example of the automatic public complaint processing device 6 in social governance, and does not constitute a limitation of the automatic public complaint processing device 6 in social governance. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, buses, etc.

[0159] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0160] In some embodiments, the memory 61 may be an internal storage unit of the automatic public complaint processing device 6 in social governance, such as a hard disk or memory of the automatic public complaint processing device 6 in social governance. In other embodiments, the memory 61 may also be an external storage device of the automatic public complaint processing device 6 in social governance, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the automatic public complaint processing device 6 in social governance. Further, the memory 61 may also include both an internal storage unit and an external storage device of the automatic public complaint processing device 6 in social governance. The memory 61 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0161] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0162] An embodiment of the present application provides a computer program product. When the computer program product runs on an automatic public complaint processing device in social governance, the automatic public complaint processing device in social governance implements the steps in any of the above-mentioned method embodiments.

[0163] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the public complaint automatic processing equipment in social governance, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, USB flash drive, mobile hard disk, disk or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0164] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0165] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0166] In the embodiments provided in the present application, it should be understood that the disclosed automatic public complaint processing system, device and method in social governance can be implemented in other ways. For example, the above-described automatic public complaint processing system and device embodiments in social governance are merely schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0167] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0168] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for automatically handling public complaints in social governance, characterized in that: include: Obtain target complaint data and complaint data packages; wherein the target complaint data is used to represent the public complaint data of a certain complainant in social governance, and the complaint data package is used to represent all historical public complaint data in social governance; Extracting features from the target complaint data to obtain target features; Determining complaint change characteristics based on the target characteristics and the complaint data packet; Based on the target complaint data and the complaint template data, determining correlation features; wherein the complaint template data includes typical complaint types and patterns summarized in social governance, and the correlation features are used to characterize the correlation between the target complaint data and the complaint template data; Performing feature fusion on the target feature, the complaint change feature and the associated feature to obtain a comprehensive feature; The comprehensive features are prioritized and classified to obtain the priority of the target complaint.

2. The method for automatically handling public complaints in social governance as claimed in claim 1, characterized in that: The feature extraction of the target complaint data to obtain the target feature includes: extracting text features from the target complaint data; Calculate the sentiment score and sentiment category of the target complaint data using a sentiment analysis model to obtain sentiment features; The text feature and the sentiment feature are integrated to obtain the target feature.

3. The method for automatically handling public complaints in social governance as claimed in claim 2, characterized in that: The determining of complaint change characteristics based on the target characteristics and the complaint data packet includes: Determining the complaint type of the target complaint according to the text features in the target features; Extracting historical complaint data of the same complaint type as the target complaint from the complaint data packet; According to the timestamps of the historical complaint data, the historical complaint data are arranged in ascending order of time to obtain a complaint time series; Based on the complaint time series, time domain features and frequency domain features are extracted, and the time domain features and the frequency domain features are integrated to obtain the complaint change features.

4. The method for automatically handling public complaints in social governance as claimed in claim 3, characterized in that: The extracting of time domain features and frequency domain features based on the complaint time series includes: Based on the complaint time series, the number of complaints in each window is calculated using a sliding window; based on the number of complaints in each window, the complaint change rate of each window is calculated; based on the change rate of each window, the time domain feature is extracted; Perform Fourier transform on the complaint time series to obtain frequency domain features.

5. The method for automatically handling public complaints in social governance as claimed in claim 1, characterized in that: The determining of the associated features based on the target complaint data and the complaint template data includes: Calculate the similarity between the target complaint data and each complaint template in the complaint template data using a similarity algorithm to obtain a similarity score for each complaint template; Selecting the complaint template with the highest similarity score from the similarity scores of each complaint template; Arrange the similarity scores of each complaint template from high to low to obtain a similarity score sequence, and extract the similarity scores of at least the first two complaint templates from the similarity score sequence; Based on the similarity score of each complaint template, calculate the variance of the similarity score and the mean of the similarity score; The similarity score of the complaint template with the highest similarity score, the similarity scores of at least the first two complaint templates in the similarity score sequence, the variance of the similarity scores and the mean of the similarity scores are integrated to obtain the associated feature.

6. The method for automatically handling public complaints in social governance as claimed in claim 1, characterized in that: The priority classification process of the comprehensive features to obtain the priority of the target complaint includes: Based on each sub-decision tree in the decision tree model, the comprehensive features are prioritized and classified to obtain sub-results of each sub-decision tree; The sub-results of each sub-decision tree are scored to obtain the score value of each category result; wherein the category result is used to represent the priority category among the sub-results of all sub-decision trees; Based on the score values ​​of each category result, the priority of the target complaint is determined.

7. The method for automatically handling public complaints in social governance as claimed in claim 6, characterized in that: The scoring process for the sub-results of each sub-decision tree to obtain the score value of each category result includes: Classify the sub-results of each sub-decision tree to obtain at least two category results; Calculating the number of sub-results of each of the at least two category results, and determining the number of sub-results of each of the at least two category results as a score value of each category result; The step of determining the priority of the target complaint based on the score value of each sub-decision tree includes: The category result with the highest score value is determined as the priority of the target complaint.

8. The method for automatically handling public complaints in social governance as claimed in claim 7, characterized in that: The scoring process for the sub-results of each sub-decision tree to obtain the score value of each category result includes: Obtaining a category score value; wherein the category score value is used to represent the score value corresponding to each category result; Selecting a score value corresponding to each category result of the at least two category results from the category score values; Calculate a score value for each category result based on the score value corresponding to each category result of the at least two category results and the number of sub-results of each category result of the at least two category results; The step of determining the priority of the target complaint based on the score value of each sub-decision tree includes: Calculate the total score of the at least two category results based on the score value of each category result in the at least two category results; Calculate an average of the at least two category results based on the total score of the at least two category results; Comparing the average of the at least two category results with the scoring threshold to obtain a comparison result; According to the comparison result, the priority of the target complaint is determined.

9. The method for automatically handling public complaints in social governance as claimed in claim 6, characterized in that: The method further comprises: Get a sample complaint dataset; Priority marking is performed on each sample complaint data in the sample complaint data set to obtain a priority label for each sample complaint data; The sample complaint data set and the priority label of each sample complaint data are divided into a training set and a test set, and the decision tree model is trained based on the training set and the test set.

10. An automatic public complaint handling system in social governance, characterized in that: include: An acquisition unit, used to acquire target complaint data and complaint data packets; wherein the target complaint data is used to represent the public complaint data of a certain complainant in social governance, and the complaint data packets are used to represent all historical public complaint data in social governance; A target feature obtaining unit, used for extracting features from the target complaint data to obtain target features; A complaint change feature determination unit, configured to determine a complaint change feature based on the target feature and the complaint data packet; A correlation feature determination unit, configured to determine a correlation feature based on the target complaint data and the complaint template data; wherein the correlation feature is used to characterize the correlation between the target complaint data and the complaint template data; A comprehensive feature obtaining unit, used for performing feature fusion on the target feature, the complaint change feature and the associated feature to obtain a comprehensive feature; The priority determination unit is used to perform priority classification processing on the comprehensive features to obtain the priority of the target complaint.

Images