Intelligent monitoring method, system and equipment for electric power service low-quality risk, and medium

Abstract

The invention discloses an intelligent monitoring method, system and device for an electric power service low-quality risk and a medium, and belongs to the technical field of electric power customer services, and the method comprises the steps: obtaining multi-source service data; based on the multi-source service data, sensitive words, complaint intention and emotion intensity characteristics in the incoming call text are extracted, and a voice risk early warning signal is generated; constructing a risk feature library of a new industry worker right and interest guarantee scene and a supplier, contractor and service provider right and interest guarantee scene, and generating a business risk early warning signal; the voice risk early warning signal and the business risk early warning signal are fused, a comprehensive risk level is determined by combining report data acquired from a customer service knowledge base in real time, and an early warning work order is generated; and starting a timed task according to the processing time limit configuration of the early warning work order, sending a processing instruction to a specified person in charge through a message pushing mode, and tracking an operation record of the early warning work order. According to the invention, the intelligent, precise and efficient level of electric power service low-cost risk monitoring is comprehensively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power customer service, in particular to an electric power service integrity risk intelligent monitoring method, system, device and medium. BACKGROUND

[0002] With the acceleration of the digital transformation of the electric power industry, using big data and artificial intelligence technology for customer service risk control has become an important means to improve power supply service quality and prevent integrity risks. By integrating marketing, customer service, production and other multi-source system data, building a risk early warning model, and accurately identifying the rights protection issues of specific groups such as "new industrial workers" and "suppliers, contractors and service providers", the risk control can be effectively transferred from post-disposal to pre-warning and intervention, and a service risk early warning system with clear management levels, convenient business processes, timely information interaction and efficient complaint disposal can be built.

[0003] However, the existing risk control mode still has the following technical problems: single risk identification dimension, which makes it difficult to achieve accurate early warning; traditional risk monitoring relies on manual experience or simple rule judgment based on work orders, and cannot effectively integrate multi-dimensional behavior data such as voice text and emotions of user calls, resulting in incomplete and untimely identification of early signs of integrity risks such as "eating, taking, card wanting" and "favored relatives and close friends"; the early warning and disposal process is fragmented, and the closed-loop control efficiency is low: the generation of risk early warning signals and subsequent work order processing, supervision, and disposal instructions lack automatic linkage, and the disposal time limit depends on manual tracking, which can easily lead to overtime and process delay, and cannot form a full-process closed-loop control of "research and judgment-tagging-verification-binding-audit-promotion". The present application solves the above problems by integrating user behavior pattern mining and intelligent business process control, and realizes a full-chain technical solution from multi-source data fusion analysis, intelligent risk fusion decision to automatic closed-loop disposal. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] To solve the above technical problems, the present application provides the following technical scheme: an electric power service integrity risk intelligent monitoring method, comprising: acquiring multi-source service data, the multi-source service data comprising call text, online customer service records, service behavior logs, work order flow information and external associated data;

[0006] Based on the multi-source service data, sensitive words, complaint intentions and emotional intensity features in the call text are extracted, and a voice risk early warning signal is generated;

[0007] Based on the work order flow information and service behavior logs, a risk feature library for new industrial worker rights protection and supplier, contractor and service provider rights protection scenarios is built, and a business risk early warning signal is generated;

[0008] Fusing the voice risk early warning signal and the business risk early warning signal, determining a comprehensive risk level in combination with the real-time obtained report data from the customer service knowledge base, and generating an early warning work order;

[0009] According to the processing time limit of the early warning work order, a start timing task is configured, a disposal instruction is sent to the designated person in charge through a message pushing mode, and the operation record of the early warning work order is tracked.

[0010] As a preferred scheme of the power service integrity risk intelligent monitoring method, the voice risk early warning signal includes:

[0011] The incoming call text and the online customer service interaction text are extracted from the multi-source service data, and text cleaning and standardization processing are performed to obtain the text data to be analyzed;

[0012] The text data to be analyzed is matched with a sensitive word library, sensitive words contained in the text are recognized, and the sensitive word occurrence frequency and position information are counted;

[0013] The text data to be analyzed is subjected to semantic analysis, and complaint intention features are extracted, the complaint intention features including negation of appeal, tendency of responsibility attribution and intensity of escalation willingness;

[0014] Based on the confidence of voice-to-text and the text sentiment analysis result, an emotion intensity feature is calculated, the emotion intensity feature being generated by anger index, anxiety index and repeated appeal frequency weighting;

[0015] The sensitive words, complaint intention features and emotion intensity features are integrated to generate a voice risk score, and a voice risk early warning signal is generated when the voice risk score exceeds a voice risk threshold.

[0016] As a preferred scheme of the power service integrity risk intelligent monitoring method, the voice risk early warning signal includes:

[0017] Work order flow information and service behavior logs related to new industrial worker rights protection and supplier contractor service provider rights protection business are extracted from the multi-source service data to generate business association data sets;

[0018] According to the event category, event type and monitoring period, the business association data sets are classified and labeled to generate labeled business risk sample data;

[0019] Based on the labeled business risk sample data, key risk features are extracted through statistical analysis, the key risk features including work order abnormal flow pattern, service response timeout frequency and repeated appeal correlation degree;

[0020] The key risk features are modeled by rules with early warning judgment objects and multi-level early warning conditions, and a risk feature library of new industry worker rights protection and supplier contractor service provider rights protection scenarios is constructed;

[0021] The real-time work order data is matched and analyzed by using the risk feature library, and when a preset business risk threshold is met, a business risk early warning signal is generated.

[0022] As a preferred scheme of the power service integrity risk intelligent monitoring method, the generation of the early warning work order comprises:

[0023] The voice risk early warning signal and the business risk early warning signal are received, and the risk type, risk level, associated work order number and trigger time information contained therein are extracted;

[0024] According to the associated work order number, it is inquired from the customer service knowledge base whether there is corresponding reported data in real time;

[0025] If the query result is that there is reported data, it is determined that the risk is a known controllable event, and no early warning work order is generated;

[0026] If the query result is that there is no reported data, the voice risk early warning signal and the business risk early warning signal are weighted and fused to calculate a comprehensive risk score;

[0027] The comprehensive risk score is compared with a preset multi-level risk threshold to determine a comprehensive risk level, and an early warning work order containing the comprehensive risk level, risk type and associated work order number is generated.

[0028] As a preferred scheme of the power service integrity risk intelligent monitoring method, the starting of the timing task according to the processing time limit of the early warning work order comprises:

[0029] According to the processing time limit and the supervision, urging and reminding strategy configured in the early warning work order, a corresponding timing monitoring task is generated;

[0030] The timing monitoring task is started, and when the reminding, urging or supervision time point is reached, a disposal instruction is sent to the designated person responsible for the early warning work order through a message pushing mode;

[0031] The early warning work order receives and records all operation behaviors in the processing process to form a whole life cycle operation record;

[0032] According to the whole life cycle operation record, the transfer, audit, archiving state and attachment operation information of the early warning work order are tracked.

[0033] As a preferred scheme of the power service integrity risk intelligent monitoring method, wherein: the statistical sensitive word appearance frequency and position information comprises:

[0034] Obtaining the sensitive word library;

[0035] The AC automatic machine algorithm is adopted to scan the to-be-analyzed text data for multiple keywords, and the sensitive words contained in the text are identified;

[0036] The starting position and the number of appearances of each identified sensitive word in the to-be-analyzed text data are recorded.

[0037] As a preferred scheme of the power service integrity risk intelligent monitoring method, wherein: the calculation of the comprehensive risk score comprises:

[0038] Obtaining the preliminary risk level of the voice risk early warning signal and the preliminary risk level of the business risk early warning signal;

[0039] The preliminary risk level of the voice risk early warning signal is weighted according to the first preset weight to obtain a voice risk weighted value;

[0040] The preliminary risk level of the business risk early warning signal is weighted according to the second preset weight to obtain a business risk weighted value, wherein the second preset weight is greater than the first preset weight;

[0041] The voice risk weighted value and the business risk weighted value are added to obtain a comprehensive risk score.

[0042] The application provides a power service integrity risk intelligent monitoring system.

[0043] To solve the above technical problems, the application further provides the following technical scheme: a power service integrity risk intelligent monitoring system, comprising: a multi-source data acquisition module for acquiring multi-source service data, wherein the multi-source service data comprises incoming call text, online customer service records, service behavior logs, work order flow information and external associated data;

[0044] A voice risk analysis module is configured to extract sensitive words, complaint intentions and emotional intensity features from the incoming call text based on the multi-source service data, and generate a voice risk early warning signal;

[0045] A business risk modeling module is configured to construct a risk feature library for new industry worker rights protection and supplier contractor service provider rights protection scenarios based on the work order flow information and the service behavior logs, and generate a business risk early warning signal;

[0046] A risk fusion decision module is configured to fuse the voice risk early warning signal and the business risk early warning signal, determine a comprehensive risk level in combination with report data acquired from a customer service knowledge base in real time, and generate an early warning work order;

[0047] An early warning closed-loop management module is configured to start a timing task according to a processing time limit of the early warning work order, send a disposal instruction to a designated person in charge through a message pushing mode, and track operation records of the early warning work order.

[0048] The application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the power service integrity risk intelligent monitoring method when executing the computer program.

[0049] The application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the power service integrity risk intelligent monitoring method.

[0050] The application has the following beneficial effects: by integrating incoming call text, online customer service records, service behavior logs, work order flow and external associated data, multi-dimensional and full-quantity data coverage of service risk is realized, laying a solid foundation for accurate identification; on this basis, sensitive words, complaint intention and emotional intensity and other multi-dimensional features are extracted from user incoming call text by using natural language processing technology, realizing fine and intelligent perception of user subjective demands, and significantly improving early warning ability of potential complaint risk; meanwhile, by constructing a risk feature library of "new industrial workers" and "supplier contractor service provider" right protection scenarios, objective behaviors such as work order abnormal flow and service overtime are modeled and analyzed, realizing active discovery of hidden integrity risk in business processes; further, by weighting and fusing voice and business risk signals and combining with report data verification, intelligent and accurate risk decision is realized, effectively avoiding false positives and resource waste; finally, by starting a timing task to drive reminder, follow-up, supervision and other disposal instructions, and tracking work order operation records throughout the process, an automatic and traceable closed-loop management system from risk discovery to disposal archiving is constructed, and the intelligent, accurate and efficient level of power service integrity risk monitoring is comprehensively improved. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0052] Figure 1 A general flowchart of an intelligent monitoring method for integrity risks of power services is provided for an embodiment of the present application.

[0053] Figure 2 A computer device diagram of an intelligent monitoring method for integrity risks of power services is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the above objectives, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0055] Embodiment 1, refer to Figure 1 For the first embodiment of the present application, the embodiment provides an intelligent monitoring method for integrity risks of power services, comprising:

[0056] S1: acquiring multi-source service data, wherein the multi-source service data comprises incoming call text, online customer service records, service behavior logs, work order flow information and external associated data;

[0057] S2: based on the multi-source service data, extracting sensitive words, complaint intentions and emotional intensity features in the incoming call text, and generating a voice risk early warning signal;

[0058] S3: based on the work order flow information and the service behavior logs, constructing a risk feature library for new industry worker rights protection and supplier contractor service provider rights protection scenarios, and generating a business risk early warning signal;

[0059] S4: fusing the voice risk early warning signal and the business risk early warning signal, combining the reported data obtained from the customer service knowledge base in real time to determine a comprehensive risk level, and generating an early warning work order;

[0060] S5: starting a timing task according to the processing time limit configuration of the early warning work order, sending a disposal instruction to a designated person of responsibility through a message pushing mode, and tracking the operation records of the early warning work order.

[0061] It should be noted that the power service integrity risk has the characteristics of strong concealment, complex inducement and rapid development, and its risk source not only includes customer complaints, service work orders and other explicit data, but also involves service personnel operation logs, external public opinion, weather warnings and other multi-dimensional implicit information. Existing risk monitoring relies on manual experience or single rule judgment, and it is difficult to automatically and accurately identify early signs of "new industrial workers" rights damage, "eat, take, card and want" and other integrity risks from massive, heterogeneous user interaction text and business flow data. At the same time, the risk early warning and subsequent disposal process often exist in disconnection, lack of full-chain automation control from risk identification, level determination, instruction issuance to work order closed loop, resulting in risk response lag and poor prevention and control effect.

[0062] Therefore, in order to solve the problems of inaccurate risk identification and fragmented early warning and disposal process, through the steps of S1-S5, first, multi-source service data is integrated to provide a comprehensive data basis for risk mining; then, risk features are independently extracted from user voice interaction and business operation logs to generate multi-source risk early warning signals, improving the comprehensiveness and accuracy of risk identification; then, by fusing and analyzing multi-source early warning signals and combining with report data for verification, comprehensive determination of risk level is realized to ensure the objectivity of early warning decision; finally, the determination result is automatically converted into executable disposal instruction, and the whole life cycle operation of the instruction is tracked to realize intelligent and automated process from risk perception to closed-loop control.

[0063] Embodiment 2, refer to Figure 1 For the second embodiment of the application, an intelligent monitoring method for power service integrity risk is provided.

[0064] S1: acquiring multi-source service data, the multi-source service data including incoming call text, online customer service record, service behavior log, work order flow information and external associated data;

[0065] Specifically, the multi-source service data is acquired by actively calling external systems through HTTP interface to obtain data. In this embodiment, the external systems include marketing management system, shared service data platform, Qiangdian Xiaoge system, online customer service system, customer service knowledge base, weather warning system and public opinion system, and a unified HTTP request protocol is configured to initiate data acquisition request to each system regularly or in real time.

[0066] For the acquisition of incoming call text and work order transfer information, the marketing management system is actively called through the HTTP interface to acquire incoming call text, work order transfer information and power failure event data. Specifically, the interface request carries an authentication token and a timestamp, and the newly added work order data within the last 5 minutes is pulled with the "incoming call time" as the screening condition, ensuring the real-time nature of data updates. The acquired data fields include call content, complainant, business type, processing status, current processing personnel and power supply unit.

[0067] For the acquisition of service behavior logs, the Qianjiang brother system and the online customer service system are actively called through the HTTP interface to obtain the operation records of service personnel and customer interactions. Specifically, trajectory information, customer feedback records and service completion status of service personnel on-site service are obtained from the Qianjiang brother system; the dialogue log of the user and the intelligent customer service, the user complaint text and the conversation end state are obtained from the online customer service system. All log data contains operation timestamp, operation personnel ID and operation type, which is used for subsequent behavior pattern analysis.

[0068] For the acquisition of external associated data, the customer service knowledge base, the weather warning system and the public opinion system are actively called through the HTTP interface. Specifically, the list of unreasonable complaint work orders reported from the customer service knowledge base is obtained for risk verification; the warning area and effective time of thunderstorms, typhoons and other adverse weather from the weather warning system; the network public opinion information related to power supply service from the public opinion system. All external data is standardized when accessed, converted to JSON format, and labeled with data source tags to generate a unified data structure of multi-source service data.

[0069] S2: Based on the multi-source service data, sensitive words, complaint intention and emotional intensity features in the incoming call text are extracted, and a voice risk early warning signal is generated;

[0070] Specifically, in order to automatically identify potential integrity risks and service complaint risks from a large amount of incoming call text, deep semantic analysis of unstructured voice text is needed. The traditional keyword matching method cannot understand the context semantics, and is prone to false positives or false negatives. Therefore, the present application adopts a hybrid analysis method combining rules and deep learning to extract multi-dimensional risk features from the text to generate accurate voice risk early warning signals. This step S2 includes steps S21 to S24.

[0071] Step S21: Extract incoming call text and online customer service interaction text from the multi-source service data, and perform text cleaning and standardization to obtain text data for analysis.

[0072] Specifically, the multi-source service data contains incoming call text converted by an automatic speech recognition (ASR) system and pure text conversation recorded by an online customer service system. The text cleaning process includes removing irrelevant characters (such as mood words like "um" and "ah"), phone numbers, user numbers, and other private information, and correcting homophonic errors. For example, "electricity bureau" and "electricity bureau" are standardized as "power supply bureau". Standardization processing also includes converting the text into a unified encoding format (such as UTF-8) and segmenting it by conversation to ensure that each unit of text data to be analyzed corresponds to a complete customer interaction process.

[0073] Step S22: Match the text data to be analyzed with the sensitive word library, identify the sensitive words contained in the text, and count the frequency and position information of the sensitive words.

[0074] Specifically, the sensitive word library is a structured database containing multiple categories of risk words, such as "complaint", "report", "eat, take, and card", "favored friends and close friends", "threat", "no benefits, no action", etc. The library supports dynamic maintenance and can be added, deleted, or modified by administrators through the "sensitive word maintenance" function. The matching process uses an efficient AC automaton algorithm that can complete multi-keyword scanning of long text within milliseconds. For the matched sensitive words, the system not only records the number of occurrences, but also records the starting and ending positions in the text for subsequent analysis of the context. For example, if the word "complaint" appears more than 3 times in a single call, it is considered a high-risk signal.

[0075] Step S23: Based on a pre-trained natural language processing model, the text data to be analyzed is subjected to semantic analysis to extract complaint intent features, including denial of appeal, responsibility attribution tendency, and escalation willingness intensity.

[0076] Specifically, a BERT pre-training model fine-tuned on power service domain corpus is used as the semantic analysis engine. The model takes the text data to be analyzed as input and outputs a high-dimensional feature vector. By connecting three independent classifiers at the top of the model, the complaint intent features in three dimensions are predicted. Denial of appeal is determined by analyzing the density and emotional polarity of negative words (such as "no", "not", "none") in the text, for example, "you have never handled" has stronger negation than "no handling". Responsibility attribution tendency is determined by identifying accusatory statements in the text that point to the power supply enterprise or its staff, for example, "it is your power supply bureau's responsibility" indicates that the user explicitly attributes the problem to the enterprise. The escalation willingness intensity is quantified by identifying "I want to find a superior" and explicitly expressing escalation statements. These three features together form a multi-dimensional complaint intent portrait.

[0077] Step S24: Based on the confidence of speech-to-text and the text sentiment analysis result, calculate the emotional intensity feature, which is generated by the anger index, the anxiety index and the repeated appeal frequency weighting;

[0078] Specifically, the speech recognition system provides a confidence score for each word when converting text, which reflects the reliability of the recognition. When the average confidence of the entire text is lower than the preset threshold (such as 0.8), it means that the user's speech speed is too fast or the user is emotional, which is a risk signal itself. At the same time, the text is classified by the emotion analysis model, and the anger index and the anxiety index are calculated. The anger index is based on the frequency and intensity of words containing "angry", "angry", "angry", etc.; the anxiety index is based on words expressing anxiety such as "when", "what", "die". In addition, the number of times the user repeats the same appeal in the call is counted. The final emotional intensity feature E is calculated by the weighted formula E = w1*anger index + w2*anxiety index + w3*repeated appeal frequency + w4*(1-average confidence), where w1, w2, w3, w4 are the emotional intensity weights, and the emotional intensity weights w1-w4 are trained according to historical data to ensure that the emotional intensity feature can effectively reflect the user's true emotional state.

[0079] Step S25: Integrate the sensitive words, complaint intention features and emotional intensity features to generate a voice risk score, and generate a voice risk warning signal when the voice risk score exceeds the voice risk threshold.

[0080] S25: Integrate the sensitive words, complaint intention features and emotional intensity features to generate a voice risk score, and generate a voice risk warning signal when the voice risk score exceeds the voice risk threshold.

[0081] Specifically, in order to quantify the multi-dimensional risk features into a unified and comparable score, the system uses a weighted summation model to calculate the voice risk score. This model ensures that different types of features (words, intentions, emotions) can be considered comprehensively, so as to generate an objective warning decision basis. The specific implementation is as follows:

[0082] Firstly, the system quantifies the "sensitive words" identified in S22. According to the preset levels in the sensitive word library, each matched sensitive word is converted into a risk score. For example, "complaint" is counted as 2 points, "report" is counted as 3 points, and "eat, take, card, want" is counted as 5 points. Then, the scores of all matched sensitive words are added up to get the total score of sensitive words. Secondly, the system quantifies the "complaint intention features" extracted in S23. This feature includes three sub-items: appeal negation, responsibility attribution tendency, and escalation willingness intensity. Each sub-item is assigned a score (e.g., 0-5) according to its severity. For example, if the text contains "you must handle it immediately, otherwise I will report", the "escalation willingness intensity" can be determined as 5 points. The scores of the three sub-items are added up to get the total score of complaint intention. Then, the system quantifies the "emotion intensity feature" calculated in S24. This feature is generated by the anger index, anxiety index, and repeated appeal frequency weighting, and its calculation result itself is a 0-10 numerical value, which can be directly used as the emotion intensity score. Finally, the system weights and fuses the above three scores. The fused voice risk score is compared with the voice risk threshold, where the voice risk threshold can be directly determined by experience.

[0083] S3: based on the work order transfer information and service behavior log, constructing a risk feature library of new industrial worker rights protection and supplier contractor service provider rights protection scene, and generating a business risk early warning signal;

[0084] Specifically, in order to accurately identify the integrity risks in the business operation process, it is necessary to deeply mine the structured work order data and unstructured service behavior logs. The traditional monitoring method based on single rule cannot effectively capture complex and hidden risk patterns. Therefore, the present application adopts a method based on multi-dimensional feature extraction and rule modeling to construct a risk feature library of a specific business scenario to generate accurate business risk early warning signals. This step S3 includes steps S31 to S35.

[0085] Step S31: extracting work order transfer information and service behavior logs related to new industrial worker rights protection and supplier contractor service provider rights protection business from the multi-source service data to form a business association data set.

[0086] Specifically, the work order flow information is derived from a marketing management system and includes a work order number, a work order type, a handling time, a processing link, a processing personnel, a transmission time, an audit state, an archiving time, and attachment information. The service behavior log is derived from a Guandian Xiaoge system and an online customer service system and includes a service personnel ID, a service time, a service location, customer feedback, operation records, and a log generation time. Through a preset business scenario identifier (such as “new industrial worker” and “supplier benefit”), work orders and service logs related to a specific scenario are filtered from the multi-source service data to form a preliminary business correlation data set.

[0087] Step S32: According to the event category, the event type, and the monitoring period, the business correlation data set is classified and labeled to generate a labeled business risk sample data.

[0088] Specifically, the event category includes “industry expansion installation”, “fault repair”, “electricity fee dispute”, and the like, and the event type is a subclass of the event category, such as “low-voltage non-residential new installation” under “industry expansion installation”. The monitoring period is a preset time window, such as “24 hours” or “7 days”. The system classifies each record in the business correlation data set according to the event category and the event type. For known risk cases, a business expert manually labels the risk level (such as high, medium, and low). For historical work orders, if they are eventually upgraded to complaints or involve “eat, take, and want” violations, they are marked as “high-risk” samples. Through this process, a batch of business risk sample data with clear risk labels for subsequent analysis is generated.

[0089] Step S33: Based on the labeled business risk sample data, key risk features are extracted through statistical analysis, including an abnormal work order flow pattern, a service response timeout frequency, and a repeated complaint correlation degree.

[0090] Specifically, the labeled business risk sample data is clustered and analyzed by association rules. First, the flow path of the work order is analyzed to identify the “abnormal flow pattern” that deviates from the standard process, such as “the work order stays in A link for too long and then jumps to C link without going through B link”. Second, the response time of each processing link is counted to calculate the “service response timeout frequency”, that is, the number of times the processing time of the work order in each link exceeds the preset SLA (service level agreement) threshold. Finally, the correlation between the work order and the user complaint is analyzed to calculate the “repeated complaint correlation degree”, that is, the proportion of multiple work orders initiated by the same user within a short period of time for the same problem. These features obtained through statistical analysis are the core elements of building a risk feature library.

[0091] Step S34: The key risk features are modeled with early warning judgment objects and multi-level early warning conditions to build a risk feature library for the new industry worker rights protection and supplier contractor service provider rights protection scenarios.

[0092] Specifically, the early warning judgment object is an entity that needs to be monitored, such as a "work order", a "service personnel", or a "power supply unit". The multi-level early warning conditions include a first-level early warning condition, a second-level early warning condition, etc., corresponding to different risk levels. The system takes the "work order abnormal flow pattern", "service response timeout frequency", and "repeated appeal correlation degree" extracted in step S33 as input features, combines them with the early warning judgment object and the multi-level early warning conditions, and forms structured rules. For example, a rule can be defined as: "when the early warning judgment object is a 'work order', and the'service response timeout frequency' is greater than 3 times, and the'repeated appeal correlation degree' is greater than 0.8, a second-level early warning is triggered". The collection of all such rules constitutes the "risk feature library" for a specific business scenario as described in the invention.

[0093] Step S35: The risk feature library is used to match and analyze real-time work order data, and a business risk early warning signal is generated when a preset business risk threshold is met.

[0094] Specifically, the system starts a timing task to periodically pull the latest work order flow information and associated service behavior logs from the work order system. These real-time data are input into the risk feature library constructed in step S34 for matching analysis. The system iterates through each rule in the feature library to calculate whether the feature values of the current work order meet the triggering conditions of any rule. For example, if the processing time of a work order in the "service scheduling" link has exceeded 2 hours, and the user has initiated 3 similar appeals within the past 24 hours, the system determines that it meets the "service response timeout frequency" and "repeated appeal correlation degree" conditions, reaches the preset business risk threshold, and immediately generates a "business risk early warning signal" carrying the work order number, risk type, risk level, and other information, which enters the subsequent early warning fusion process.

[0095] Specifically, the preset business risk threshold is determined based on historical risk event data and expert experience, and is used to determine whether the current work order constitutes a business risk. The setting of this threshold ensures the sensitivity and accuracy of risk early warning, avoiding the situation of missed reports due to a too high threshold or false reports due to a too low threshold. The determination process follows the following steps:

[0096] Extract samples of work orders confirmed as "high risk" within the past two years from the historical work order database, which involve integrity issues such as "new industrial worker" rights violations or "supplier contractor service provider" complaints. Perform cluster analysis on these samples and calculate the statistical distribution of their key risk characteristics, such as the average value of "service response timeout frequency" is 4.2 times, and the standard deviation is 1.1 times; the median of "repeated appeal correlation degree" is 0.75. Combine the judgment of business experts to modify and qualify the statistical results. For example, experts believe that when the "work order abnormal flow pattern" is triggered (i.e., the work order skips key audit links), it should be considered high risk regardless of other indicators. Therefore, triggering such a pattern is directly defined as reaching the highest risk threshold. Quantify multiple risk characteristics into a comprehensive score through a weighted scoring model, and set threshold values. For example, the comprehensive score calculation formula is: comprehensive score = 2 * abnormal flow pattern (yes = 1, no = 0) + 1.5 * timeout frequency + 1.0 * repeated appeal correlation degree. According to the distribution of historical data and expert opinions, the first-level warning threshold is set to 8 points, the second-level warning threshold is set to 6 points, and the third-level warning threshold is set to 4 points. Therefore, when the real-time work order data is matched and analyzed by the risk characteristic library, if the calculated comprehensive score is equal to or exceeds the preset business risk threshold (such as 6 points), the system generates a business risk warning signal of the corresponding level.

[0097] S4: Fuse the voice risk warning signal and the business risk warning signal, determine the comprehensive risk level in combination with the reported data obtained from the customer service knowledge base in real time, and generate a warning work order;

[0098] Specifically, in order to prevent repeated warnings on known and reported repeated events, causing resource waste and information overload, the present application introduces a verification and fusion decision mechanism based on reported data before generating a warning work order. This mechanism makes a comprehensive evaluation of multiple sources of risk signals and makes a final decision in combination with external reported information, ensuring the accuracy and necessity of the warning. This step S4 includes steps S41 to S45.

[0099] Step S41: Receive the voice risk warning signal and the business risk warning signal, and extract the risk type, risk level, associated work order number, and trigger time information contained therein.

[0100] Specifically, after receiving the voice risk warning signal generated by S2 and the business risk warning signal generated by S3, the system first parses the structured data package of the signal. Each warning signal contains a unique identifier, risk type (such as "new industrial worker rights" and "supplier complaint"), preliminary determined risk level (such as first level and second level), associated original work order number, and accurate timestamp of signal generation. The system takes these information as the basis for subsequent fusion analysis.

[0101] Step S42: According to the associated work order number, query whether there is corresponding reported data in the customer service knowledge base in real time.

[0102] Specifically, the system takes the associated work order number as the primary key and initiates a real-time query request to the customer service knowledge base through an HTTP interface. The purpose of the query is to confirm whether the problem reflected by the work order has been marked as "reported" or "known problem" in the knowledge base. For example, due to weather warnings, large-scale power outages may have been reported in advance. The query process requires completion within milliseconds to ensure the real-time nature of the entire warning process.

[0103] Step S43: If the query result is that there is reported data, it is determined that the risk is a known controllable event, and no warning work order is generated.

[0104] Specifically, when the system obtains the report record matching the associated work order number from the response of the customer service knowledge base, the system immediately executes this step. The logic of this step is that since the problem has been known and documented by the superior or relevant department in advance, the event is under control and there is no need to trigger the warning process again. The system will record the log of this "reporting to suppress warning" for subsequent process audit and optimization, but will not execute the operation of generating a warning work order.

[0105] Step S44: If the query result is that there is no reported data, the voice risk warning signal and the business risk warning signal are weighted and fused to calculate the comprehensive risk score.

[0106] Specifically, when the system confirms that there is no corresponding reported data for the associated work order number, the weighted fusion calculation process is started to generate an objective and quantitative risk assessment result. This process takes into account both user subjective expression and objective business behavior, avoiding the one-sidedness of a single signal source. The specific implementation is as follows:

[0107] First, the system extracts the "preliminary risk level" from the voice risk warning signal to be fused. This level is determined by comparing the voice risk score calculated based on sensitive words, complaint intent, and emotional intensity in step S2 with a preset threshold, for example, "level two". Similarly, the "preliminary risk level" is extracted from the business risk warning signal, for example, "level one".

[0108] Then, the system performs a weighting calculation. The system converts the preliminary risk level of the voice risk early warning signal into a numerical value (e.g., level 1 = 4 points, level 2 = 3 points, level 3 = 2 points, and level 4 = 1 point) according to a pre-configured weight rule, and then multiplies the first preset weight (e.g., 0.3) to obtain a voice risk weighted value. Meanwhile, after converting the preliminary risk level of the business risk early warning signal into a corresponding numerical value, the second preset weight (e.g., 0.7) is multiplied to obtain a business risk weighted value. This design reflects the principle that the "business risk" has a higher priority in the comprehensive determination, that is, the second preset weight is greater than the first preset weight.

[0109] Finally, the system adds the calculated voice risk weighted value and the business risk weighted value to obtain a final comprehensive risk score. For example, after a level 2 voice risk (3 points * 0.3 = 0.9) is fused with a level 1 business risk (4 points * 0.7 = 2.8), the comprehensive risk score is 3.7. This comprehensive risk score serves as the core basis for subsequent determination of a comprehensive risk level and generation of an early warning work order.

[0110] Step S45: Comparing the comprehensive risk score with a pre-set multi-level risk threshold value, determining a comprehensive risk level, and generating an early warning work order containing the comprehensive risk level, risk type, and associated work order number.

[0111] Specifically, the system maintains a risk level mapping table, and the pre-set multi-level risk threshold value is, for example, level 1 early warning for a comprehensive score ≥ 90 points, level 2 early warning for 80 ≤ score < 90, and level 3 early warning for 70 ≤ score < 80. The comprehensive risk score (e.g., 82 points) calculated in step S44 is compared with the mapping table to determine the final comprehensive risk level (level 2). Subsequently, the system calls a work order generation service to create a new early warning work order. The core information of the early warning work order includes a unique early warning work order number automatically generated by the system, the determined comprehensive risk level, the comprehensive risk type, the original associated work order number, and a label of "automatically initiated". Finally, the early warning work order is stored in a database and enters a to-do process.

[0112] S5: Starting a timing task according to the processing time limit configuration of the early warning work order, sending a disposal instruction to a designated person in charge through a message push method, and tracking the operation record of the early warning work order.

[0113] Specifically, in order to ensure that the early warning work order is processed in a timely and effective manner, the system establishes a closed-loop management and control mechanism from instruction issuance to state tracking. This mechanism drives reminder, follow-up and supervision processes through automatic timing tasks, and records the whole life cycle operation of the work order to ensure that the risk disposal process is traceable and auditable. This step S5 includes steps S51 to S54.

[0114] Step S51: According to the processing time limit and supervision, urging, and reminding strategies configured in the early warning work order, a corresponding timing monitoring task is generated.

[0115] Specifically, the early warning work order contains the processing time limit rule configured by the "early warning business library management" in its metadata when it is generated. For example, a secondary early warning work order requires that the closed loop be completed within 12 hours. According to this rule, the system automatically creates a timing monitoring task. The task is divided into multiple subtasks: a "reminder" task is triggered 20 minutes after the work order is generated (1 / 36 of the total time limit), a "urging" task is triggered when the work order has existed for 10 hours (5 / 6 of the total time limit), and a "supervision" task is triggered when the work order exceeds 12 hours. Each subtask is associated with an accurate execution time point.

[0116] Step S52: The timing monitoring task is started, and when the preset reminder, urging, or supervision time point is reached, a disposal instruction is sent to the designated person responsible for the early warning work order through message pushing.

[0117] Specifically, the system polls all pending timing tasks. When the system time reaches the preset time point of a certain task, the message sending process is triggered. The system determines the designated person responsible for the current link according to the "early warning process role configuration", and generates a specific disposal instruction according to the template preset in the "message template configuration" for the "reminder", "urging", "supervision" and other scenarios. The instruction is sent through the "message pushing method", which can be short message, elink message or intelligent external call according to the configuration. For example, for "reminder", the system calls the short message service to send the instruction to the mobile phone of the person responsible; for "supervision", the system may call the intelligent external call service to make a phone call.

[0118] Step S53: Receive and record all operation behaviors of the early warning work order during the processing process to form a full life cycle operation record.

[0119] Specifically, whenever the designated person responsible for the early warning work order processes it, the system will record the operation. For example, when the person responsible for the work order queries the work order, adds a note, uploads an attachment, or performs operations such as "transfer", "backtracking", "archive" and the like through the "work order proxy processing" function, the service in the background of the system will capture these behaviors. Each operation record contains the operation timestamp, the operator ID, the operation type (such as "uploading an attachment"), the operation content summary (such as "uploading a scene photo.jpg"), and the work order state before and after the operation. All this information is written into the database in real time to form a complete operation log of the early warning work order.

[0120] Step S54: According to the full life cycle operation record, the transfer, audit, and archive state of the early warning work order and the attachment operation information are tracked to realize closed loop control of service risk.

[0121] Specifically, by querying and analyzing the full lifecycle operation records generated in step S53, the system can reconstruct the processing path of early warning work orders in real time. For example, the system can determine whether the work order has been transferred from the "provincial service dispatcher" to the "municipal service dispatcher," whether the "supervised work order review" has been completed, and whether it has finally been archived by the "supervised work order archiving" function. Simultaneously, the system can track the upload, viewing, and download records of all attachments. This tracking information is not only used for the "supervision status display" and "expedited status display" on the "homepage dashboard," but also provides data support for functions such as "expedited status query" and "abnormal risk details query," ultimately achieving a complete closed loop from risk discovery to handling.

[0122] Example 3 is the third embodiment of the present invention. This embodiment provides an intelligent monitoring system for corruption risks in power services, including: a multi-source data acquisition module for acquiring multi-source service data, wherein the multi-source service data includes incoming call text, online customer service records, service behavior logs, work order flow information and external related data;

[0123] The voice risk analysis module is used to extract sensitive words, complaint intent and emotional intensity features from incoming call text based on the multi-source service data, and generate voice risk warning signals.

[0124] The business risk modeling module is used to construct a risk feature library for the protection of the rights and interests of new industrial workers and suppliers, contractors and service providers based on the work order flow information and service behavior logs, and generate business risk early warning signals.

[0125] The risk fusion decision module is used to fuse the voice risk warning signal and the business risk warning signal, combine the reported data obtained in real time from the customer service knowledge base to determine the comprehensive risk level, and generate a warning work order;

[0126] The early warning closed-loop management module is used to configure and start a timed task according to the processing time limit of the early warning work order, send the handling instructions to the designated person in charge through message push, and track the operation record of the early warning work order.

[0127] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the intelligent monitoring method for corruption risks in power services.

[0128] Example 4, refer to Figure 2For the fourth embodiment of the present application, which is different from the first three embodiments, the function, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or the part of the technical solution that essentially contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0129] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or in conjunction with these instructions execution systems, apparatuses, or devices. For the purpose of this specification, the "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport programs for use by an instruction execution system, apparatus, or device, or in conjunction with these instruction execution systems, apparatuses, or devices.

[0130] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, as the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by electronic editing, interpretation, or necessary processing, and then stored in a computer memory if necessary. Other suitable media can also be used.

[0131] It should be understood that portions of the present application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, implementation can be with any or a combination of the following technologies, which are all well known in the art: a discrete logic circuit having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0132] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. An intelligent monitoring method for integrity risk of electric power service, characterized in that: The method comprises the steps of: obtaining multi-source service data, including incoming call text, online customer service records, service behavior logs, work order transfer information, and external associated data; based on the multi-source service data, extracting sensitive words, complaint intent and emotional intensity features from the incoming call text, and generating a voice risk early warning signal; based on the work order transfer information and service behavior logs, a risk feature library for new industry worker rights protection and supplier contractor service provider rights protection scenarios is constructed, and a business risk early warning signal is generated; fuse the voice risk early warning signal and the business risk early warning signal, combine the real-time report data obtained from the customer service knowledge base to determine the comprehensive risk level, and generate a warning work order; according to the processing time limit configuration of the warning work order, start a timing task, send a disposal instruction to the designated person of responsibility through message pushing, and track the operation record of the warning work order.

2. The method of claim 1, wherein the method further comprises: The method comprises the steps of: extracting incoming call text and online customer service interaction text from the multi-source service data, and performing text cleaning and standardization processing to obtain text data for analysis; matching the text data for analysis with a sensitive word library to identify sensitive words contained in the text and count the frequency and position information of the sensitive words; performing semantic analysis on the text data for analysis to extract complaint intent features, including negation, responsibility attribution tendency and escalation willingness intensity; based on the confidence of voice to text and the text sentiment analysis result, calculate the emotional intensity feature, which is generated by anger index, anxiety index and repeated complaint frequency weighting; integrate the sensitive words, complaint intent features and emotional intensity features to generate a voice risk score, and generate a voice risk early warning signal when the voice risk score exceeds the voice risk threshold.

3. The method of claim 2, wherein the method further comprises: The method comprises the steps of: extracting work order transfer information and service behavior logs related to new industry worker rights protection and supplier contractor service provider rights protection business from the multi-source service data to generate business associated data set; according to the event category, event type and monitoring period, classify and label the business associated data set to generate labeled business risk sample data; based on the labeled business risk sample data, extract key risk features through statistical analysis, including work order abnormal transfer mode, service response timeout frequency and repeated complaint correlation degree; rule-based modeling of the key risk features, early warning judgment objects and multi-level early warning conditions to construct a risk feature library for new industry worker rights protection and supplier contractor service provider rights protection scenarios; use the risk feature library to perform matching analysis on real-time work order data, and generate a business risk early warning signal when the preset business risk threshold is met.

4. The method of claim 3, wherein the method further comprises: The method comprises the steps of: receiving the voice risk early warning signal and the business risk early warning signal, extracting the risk type, risk level, associated work order number and trigger time information contained therein; according to the associated work order number, query whether there is corresponding reported data in the customer service knowledge base in real time; If the query result is that there is already reported data, it is determined that the risk is a known controllable event, and no early warning work order is generated; If the query result is that there is no already reported data, the voice risk early warning signal and the business risk early warning signal are weighted and fused to calculate a comprehensive risk score; The comprehensive risk score is compared with a preset multi-level risk threshold to determine a comprehensive risk level, and an early warning work order containing the comprehensive risk level, risk type and associated work order number is generated.

5. The method of claim 4, wherein the method further comprises: The processing time limit configuration of the early warning work order includes: According to the processing time limit and the supervision, urging and reminding strategy configured in the early warning work order, corresponding timing monitoring tasks are generated; The timing monitoring tasks are started, and when the reminding, urging or supervision time point is reached, disposal instructions are sent to the designated person responsible for the early warning work order through message pushing; All operation behaviors of the early warning work order in the processing process are received and recorded to form a whole life cycle operation record; According to the whole life cycle operation record, the transfer, audit, archiving state and attachment operation information of the early warning work order are tracked.

6. The method of claim 5, wherein the method further comprises: The statistical sensitive word frequency and position information includes: Obtaining the sensitive word library; Using the AC automatic machine algorithm to scan the to-be-analyzed text data for multiple keywords to identify the sensitive words contained in the text; The starting position and occurrence frequency of each identified sensitive word in the to-be-analyzed text data are recorded.

7. The method of claim 6, wherein the method further comprises: The calculation of the comprehensive risk score includes: Obtaining the preliminary risk level of the voice risk early warning signal and the preliminary risk level of the business risk early warning signal; The preliminary risk level of the voice risk early warning signal is weighted according to a first preset weight to obtain a voice risk weighted value; The preliminary risk level of the business risk early warning signal is weighted according to a second preset weight to obtain a business risk weighted value, wherein the second preset weight is greater than the first preset weight; The voice risk weighted value and the business risk weighted value are added to obtain a comprehensive risk score.

8. An intelligent monitoring system for integrity risks in power service, applying an intelligent monitoring method for integrity risks in power service according to any one of claims 1-7, characterized in that, It includes: A multi-source data acquisition module for acquiring multi-source service data, the multi-source service data including incoming call text, online customer service records, service behavior logs, work order flow information and external associated data; A voice risk analysis module for extracting sensitive words, complaint intent and emotional intensity features from the incoming call text based on the multi-source service data to generate a voice risk early warning signal; A business risk modeling module for constructing a risk feature library for new industry worker rights protection and supplier contractor service provider rights protection scenarios based on the work order flow information and service behavior logs to generate a business risk early warning signal; A risk fusion decision module for fusing the voice risk early warning signal and the business risk early warning signal, combining the real-time reported data obtained from the customer service knowledge base to determine a comprehensive risk level, and generating an early warning work order; An early warning closed-loop management module for starting timing tasks according to the processing time limit of the early warning work order, sending disposal instructions to designated persons responsible through message pushing, and tracking the operation record of the early warning work order. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The computer program is executed by the processor to implement the steps of the power service integrity risk intelligent monitoring method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the power service integrity risk intelligent monitoring method in any one of claims 1 to 7.

Images