Customer complaint processing method and device

Abstract

The invention provides a customer complaint processing method and device, which can improve the accuracy and efficiency of customer complaint processing. The customer complaint processing method comprises the following steps: acquiring customer complaint information; responding to the customer complaint information, and obtaining customer information and network fault information; calculating a complaint emergency degree score, a network real-time state score and a customer value score based on the customer information and the network fault information; calculating a comprehensive score according to the complaint emergency degree score, the network real-time state score and the customer value score; and network resource allocation optimization is carried out based on the comprehensive score.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and telecommunications service operation technology, and in particular to a customer complaint handling method and apparatus. Background Technology

[0002] In today's digital age, telecommunications services are booming, with their scope constantly expanding to encompass voice calls, data transmission, multimedia entertainment, and many other areas, greatly enriching people's daily lives and work. Simultaneously, with the popularization of communication technology and the gradual decrease in the price of communication equipment, the number of telecommunications customers has increased dramatically, forming a massive customer base. However, in the process of rapid development of telecommunications services, customer complaints have become increasingly prominent. Due to the complexity of communication systems, the dynamic changes in the network environment, and the diversification of customer needs, the types and number of customer complaints have become increasingly numerous. The sharp rise in the number of complaints has brought enormous pressure and challenges to the customer complaint handling work of telecommunications operators. Traditional customer complaint handling methods mainly rely on manual operation. This handling model has gradually exposed many drawbacks when faced with increasingly complex and numerous customer complaints, seriously affecting the efficiency and quality of customer complaint handling. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention is proposed. Embodiments of this invention provide a customer complaint handling method and apparatus, which can improve the accuracy and efficiency of customer complaint handling.

[0004] According to one aspect of the present invention, a customer complaint handling method is provided, comprising: acquiring customer complaint information; in response to the customer complaint information, acquiring customer information and network fault information; calculating a complaint urgency score, a network real-time status score, and a customer value score based on the customer information and the network fault information; calculating a comprehensive score based on the complaint urgency score, the network real-time status score, and the customer value score; and optimizing network resource allocation based on the comprehensive score.

[0005] In one embodiment, the network fault information includes the number of affected customers and the duration of the fault, and the customer information includes the customer service level agreement level; wherein, calculating the complaint urgency score includes: calculating the number of affected customers multiplied by a first weight, the customer service level agreement level multiplied by a second weight, and the fault duration multiplied by a third weight; and adding the product of the number of affected customers multiplied by the first weight, the product of the customer service level agreement level multiplied by the second weight, and the product of the fault duration multiplied by the third weight to calculate the complaint urgency score.

[0006] In one embodiment, the network fault information includes the number of alarms and the alarm level; The calculation of the real-time network status score includes: obtaining the number and level of alarms across multiple dimensions; wherein each dimension corresponds to at least one alarm level and number of alarms; processing the alarm level and number of alarms for each dimension according to a preset scoring function to obtain a score value for each dimension; the scoring function is a linear transformation function or a nonlinear transformation function; weighting and summing the score values ​​for each dimension based on preset weights to obtain a weighted summation result; the weights satisfy a value range between 0 and 1 and the sum of all weights is 1; and calculating the real-time network status score based on the weighted summation result.

[0007] In one embodiment, the customer information includes average revenue per customer, customer online time, customer contract status, and customer historical complaint data; calculating the customer value score includes: using the average revenue per customer, the customer online time, the customer contract status, and the customer historical complaint data as hyperparameters of a preset model; inputting customer information and network failure information into the preset model to predict customer churn risk value; and using the customer churn risk value to inversely map to the customer value score.

[0008] In one embodiment, the preset model is trained using at least one of random forest, logistic regression, and XGBoost algorithms.

[0009] In one embodiment, in response to customer complaint information, obtaining customer information and network fault information includes: preprocessing the customer complaint information and inputting it into a Chinese text analysis model; the Chinese text analysis model is obtained by adjusting the complaint data containing fault labels based on a pre-trained Chinese language model; processing the customer complaint information through the Chinese text analysis model and outputting at least one classification label for indicating a fault domain; determining the management department corresponding to the fault domain classification label according to a predefined mapping relationship, thereby generating network fault information containing the fault domain and the corresponding management department.

[0010] In one embodiment, a comprehensive score is calculated based on the complaint urgency score, the network real-time status score, and the customer value score, including: calculating a comprehensive score of the routing result based on the complaint urgency score, the network real-time status score, the customer value score, and preset weight coefficients; wherein the weight coefficients include at least a first weight corresponding to the urgency score, a second weight corresponding to the network real-time status score, and a third weight corresponding to the customer value score.

[0011] In one embodiment, optimizing network resource allocation based on the comprehensive score includes: comparing the comprehensive score of the routing result with a preset threshold; when the comprehensive score of the routing result is greater than the preset threshold, determining the management department corresponding to the routing result greater than the preset threshold as the department to be routed; selecting the management department with the highest comprehensive score among the departments to be routed as the target routing department; and using the target routing department to process the customer complaint information.

[0012] In one embodiment, optimizing network resource allocation based on the comprehensive score further includes: when the comprehensive score of the routing result is not greater than the preset threshold, determining the routing result as a result to be dispatched; wherein the result to be dispatched is subject to secondary review and dispatch by human customer service.

[0013] According to another aspect of the present invention, a customer complaint processing apparatus is provided, comprising: an acquisition module for acquiring customer complaint information; a response module for acquiring customer information and network fault information in response to the customer complaint information; a calculation module for calculating a complaint urgency score, a network real-time status score, and a customer value score based on the customer information and the network fault information; a scoring module for calculating a comprehensive score based on the complaint urgency score, the network real-time status score, and the customer value score; and an optimization module for optimizing network resource allocation based on the comprehensive score.

[0014] The customer complaint handling method and apparatus provided by this invention analyzes the urgency and type of customer complaints from three dimensions: complaint urgency score, network real-time status score, and customer value score, forming a comprehensive score. Combining the comprehensive score with network resource allocation optimization can improve complaint handling efficiency, accurately locate the causes of complex complaints, and optimize dynamic resource scheduling, thereby improving customer service experience and reducing customer churn rate. Attached Figure Description

[0015] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same parts or steps.

[0016] Figure 1 This is a flowchart illustrating a customer complaint handling method provided in an exemplary embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the structure of a customer complaint processing device provided in an exemplary embodiment of the present invention. Detailed Implementation

[0018] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0019] With the rapid development of telecommunications services and the dramatic expansion of the customer base, customer complaints have become increasingly complex and numerous. In today's highly competitive market environment, efficient and accurate handling of customer complaints is crucial for improving customer experience and maintaining operator competitiveness. However, existing customer complaint handling systems still rely primarily on manual intervention and static rules, exhibiting significant shortcomings in processing efficiency, resource allocation, and emergency response, making them ill-suited to the dynamic and complex modern communication network environment.

[0020] Specifically, the existing technology mainly has the following problems: First, routing decisions are inefficient and inaccurate. Traditional complaint ticket routing relies heavily on human agents reading, understanding, and judging the complaint content, or simply calculating weights based on a few preset, fixed dimensions (such as business type). This approach fails to comprehensively consider the urgency inherent in the complaint, the real-time network status of the current area, and dynamic information such as customer value, resulting in a slow routing decision process, with an average routing time of 2 to 5 minutes per ticket. More importantly, due to the lack of intelligent analysis, the routing error rate remains high, reaching as high as 15% according to practical statistics, causing tickets to be ineffectively circulated between departments and significantly delaying problem resolution.

[0021] Second, resource scheduling is rigid and unable to handle sudden traffic surges. The routing mechanism, based on fixed rules, lacks flexibility and cannot perceive or adapt to dynamic network changes. When local network failures occur or a surge in complaints occurs in a specific area, the system still directs a large number of work orders to overloaded processing departments according to established rules, failing to dynamically schedule tasks to relatively idle departments. This rigid scheduling model leads to excessive local resource load; during peak business hours, approximately 20% of customer complaints require more than two hours to be initially processed, severely impacting customer satisfaction.

[0022] Third, emergency response is severely lagging and automation is insufficient. When faced with large-scale, concentrated complaints triggered by sudden network failures (such as base station outages or severed fiber optic cables), the existing system lacks autonomous detection and emergency response capabilities. Maintenance personnel must manually discover the fault, analyze its impact, and then manually intervene to adjust the system's routing strategy. This process is excessively time-consuming, preventing the complaint handling system from coordinating with the network fault self-healing system at critical moments, missing the optimal window for emergency response, and amplifying the negative impact of the fault on customer businesses.

[0023] Therefore, there is an urgent need in this field for a technical solution that can intelligently analyze complaint content, dynamically perceive network status, and automatically make optimal routing decisions to overcome the aforementioned drawbacks of existing technologies.

[0024] To solve the above problems, Figure 1 This is a flowchart illustrating a customer complaint handling method provided in an exemplary embodiment of the present invention, as shown below. Figure 1 As shown, firstly, obtain customer complaint information (see...). Figure 1 (S110), and secondly, in response to customer complaint information, obtain customer information and network failure information (see S110). Figure 1 (S120), and then, based on customer information and network failure information, calculate the complaint urgency score, the real-time network status score, and the customer value score (see S120). Figure 1 (S130), then, based on the complaint urgency score, network real-time status score, and customer value score, a comprehensive score is calculated (see S130). Figure 1 (S140). Finally, network resource allocation is optimized based on the comprehensive score (see S140). Figure 1 (S150).

[0025] The following text combines Figure 1 This application provides a more detailed description of the customer complaint handling method provided in the embodiments.

[0026] In S110, obtain customer complaint information.

[0027] In some embodiments, the system can collect data in real time from multiple systems, such as CRM (Customer Relationship Management) systems, network monitoring systems, and customer interaction systems, as data sources for subsequent analysis. For example, it can collect structured data, including customer profiles, work order history, package information, account balances, and billing information from the CRM system. It can also collect unstructured data, such as voice complaints converted to text and online chat logs, and dynamic data, such as network resource data, network alarms, and network performance KPIs. When a customer complaint is received, various relevant information about the customer who initiated the complaint can be extracted to prepare for subsequent analysis.

[0028] In S120, in response to customer complaint information, customer information and network failure information are obtained.

[0029] In some embodiments, customer complaint information is preprocessed and input into a Chinese text analysis model; the Chinese text analysis model is obtained by adjusting the complaint data containing fault labels based on a pre-trained Chinese language model; the customer complaint information is processed by the Chinese text analysis model to output at least one classification label for indicating the fault domain; according to a predefined mapping relationship, the management department corresponding to the fault domain classification label is determined, thereby generating network fault information containing the fault domain and the corresponding management department.

[0030] For example, the BERT-wwm (Whole Word Masking for Chinese BERT, a pre-trained Chinese language model) model can be introduced to optimize the accuracy of complaint text analysis and fault reasoning, providing reliable semantic feature input. BERT-wwm accurately identifies complete fault entities in complaints through whole-word masking and dynamic phrase-level masking, improving semantic understanding capabilities, such as identifying "poor 5G signal" instead of just "5G" or "signal," and can also improve the accuracy of sentiment analysis, such as training "very dissatisfied" as a whole phrase.

[0031] In S130, based on customer information and network failure information, a complaint urgency score, a real-time network status score, and a customer value score are calculated.

[0032] As one possible approach, the urgency score of a complaint can be calculated using the number of affected customers, the duration of the outage, and the service-level agreement (SLA) level as evaluation metrics. The number of affected customers (users) represents the number of customers impacted by the same outage; for example, analyzing customer complaints can clarify whether a base station outage affects 1000 people or a single household's broadband connection. The outage duration (Duration) can be, for example, 2 hours or just 5 minutes after it occurred. The service-level agreement (SLA) level can be categorized as Gold, Silver, or Standard, etc.

[0033] In some embodiments, calculating the urgency score of a complaint includes: multiplying the number of affected customers by a first weight, the customer service level agreement level by a second weight, and the duration of the outage by a third weight; and summing the products of the number of affected customers multiplied by the first weight, the customer service level agreement level by the second weight, and the duration of the outage by the third weight to calculate the urgency score of the complaint.

[0034] The urgency score of a complaint is calculated using the following formula: S(urgency)=w1×N+ w2×L+ w3×T Formula 1; In Formula 1, N represents the number of affected customers, L represents the customer service level agreement level, T represents the duration of the outage, w1 represents the weighting coefficient for the number of affected customers, w2 represents the weighting coefficient for the customer service level agreement level, and w3 represents the weighting coefficient for the duration of the outage. Based on historical complaint tickets, a training push is performed every hour to update the weights of each parameter in the complaint urgency model. Through the complaint urgency model, the complaint urgency score S of the complaint ticket is output. That is to say, each weighting coefficient can be dynamically adjusted, determined by the model calculation, or manually adjusted. The value of the customer service level agreement level can be pre-assigned, for example, a value of 5 for gold customers and a value of 1 for ordinary customers in the calculation.

[0035] In some embodiments, network fault information includes the number of alarms and alarm levels; wherein, calculating the real-time network status score includes: obtaining the number of alarms and alarm levels for multiple dimensions; wherein each dimension corresponds to at least one alarm level and alarm number; processing the alarm level and alarm number for each dimension according to a preset scoring function to obtain a score value for each dimension; the scoring function is a linear transformation function or a nonlinear transformation function; weighting and summing the score values ​​for each dimension based on preset weights to obtain a weighted summation result; the weights satisfy a value range between 0 and 1 and the sum of all weights is 1; calculating the real-time network status score based on the weighted summation result.

[0036] As one possible implementation, a weighted summation model is used, and the real-time network status score is calculated using Formula 2 based on network evaluation metrics (such as the number of alarms and alarm levels): S(network) = 100−i=1∑nw i ⋅f i Formula 2: (Alarm Level, Number of Alarms) In Formula 2, w i w represents the weight of the i-th dimension, 0 ≤ w i ≤ 1, and ∑w i =1, f i This represents the scoring function for the i-th dimension (usually a linear or non-linear transformation). It can be understood that the dimension here can be interpreted as the alarm group, grouping alarms of the same type or in the same region into one dimension, calculating a unified network real-time status score, and then weighted summing the network real-time status scores from different dimensions to evaluate the overall network real-time status score.

[0037] In some embodiments, customer information includes average revenue per user (ARPU), customer engagement time, customer contract status, and historical customer complaint data. ARPU represents the average revenue a business earns from each user over a given period and is typically used to measure a business's revenue-generating capacity and customer value. Customer engagement time includes the duration of customer interaction and spending trends; customer contract status includes whether the contract is still in effect and whether any penalties apply for breach of contract; and historical customer complaint data includes whether the customer has a history of complaints.

[0038] Customer information, including average revenue per customer, customer engagement time, customer contract status, and historical customer complaint data, is used as hyperparameters in a pre-defined model to output a customer value score. For example, customer information is acquired, including multiple customer samples and feature data for each sample. From the feature data, features related to average revenue per user, customer engagement time, customer contract status, and historical customer complaint data are extracted as model input features. These input features are then fed into a pre-trained prediction model, which outputs a customer churn risk value. This risk value is then inversely mapped to a priority-weighted customer value score.

[0039] In some embodiments, the prediction model is trained using training samples on at least one of the algorithms Random Forest (RF), Logistic Regression (LR), and XGBoost. The algorithm can be selected based on the data structure and actual needs.

[0040] In S140, a comprehensive score for the routing results is calculated based on the complaint urgency score, the network real-time status score, the customer value score, and preset weight coefficients. The weight coefficients include at least a first weight corresponding to the urgency score, a second weight corresponding to the network real-time status score, and a third weight corresponding to the customer value score.

[0041] In some embodiments, the first, second, and third weights are set according to the user's actual needs. For example, if the user's urgency is higher, the first weight is given a larger proportion; if the user's need for real-time network status is higher, the second weight is given a larger proportion. Furthermore, the first, second, and third weights can also be provided by the system based on a data-trained model, or they can be manually adjusted. For example, if the first weight is preset to 20%, the second weight to 30%, and the third weight to 50%, and the user feedback is unsatisfactory, the weight allocation can be adjusted based on user feedback to achieve closed-loop regulation.

[0042] As one possible implementation method, a comprehensive score can be calculated using Formula 3: Score=W(urgency)×S(urgency) + W(network)×S(network) + W(customer)×V(customer) Formula 3; Among them, Score represents the overall score, W (urgency) represents the first weight, W (network) represents the second weight, W (customer) represents the third weight, S (urgency) represents the complaint urgency score, S (network) represents the network real-time status score, and V (customer) represents the customer value score.

[0043] In S150, network resource allocation is optimized based on comprehensive scoring.

[0044] For example, the overall score of the routing results is compared with a preset threshold; when the overall score of the routing results is greater than the preset threshold, the management department corresponding to the routing results that are greater than the preset threshold is identified as the department to be routed; the management department with the highest overall score among the departments to be routed is identified as the target routing department; the target routing department is used to handle customer complaint information.

[0045] When the overall score of the routing result is not greater than the preset threshold, the routing result is determined as a result to be dispatched. The result to be dispatched is then reviewed and dispatched by human customer service to ensure the service quality of low confidence decision paths.

[0046] Figure 2 This is a schematic diagram of the structure of a customer complaint processing device provided in an exemplary embodiment of the present invention, as shown below. Figure 2 As shown, the customer complaint handling device 2 includes: an acquisition module 21 for acquiring customer complaint information; a response module 22 for responding to customer complaint information and acquiring customer information and network fault information; a calculation module 23 for calculating a complaint urgency score, a network real-time status score, and a customer value score based on the customer information and network fault information; a scoring module 24 for calculating a comprehensive score based on the complaint urgency score, the network real-time status score, and the customer value score; and an optimization module 25 for optimizing network resource allocation based on the comprehensive score.

[0047] In one embodiment, the network fault information includes the number of affected customers and the duration of the fault, and the customer information includes the customer service level agreement level; wherein, the calculation module 23 can be configured to: calculate the number of affected customers multiplied by a first weight, the customer service level agreement level multiplied by a second weight, and the fault duration multiplied by a third weight; and add the product of the number of affected customers multiplied by the first weight, the product of the customer service level agreement level multiplied by the second weight, and the product of the fault duration multiplied by the third weight to calculate the urgency score of the complaint.

[0048] In one embodiment, network fault information includes the number of alarms and alarm levels; wherein, the calculation module 23 can be configured to: acquire the number of alarms and alarm levels of multiple dimensions; wherein, each dimension corresponds to at least one alarm level and alarm number; process the alarm level and alarm number of each dimension according to a preset scoring function to obtain a score value for each dimension; the scoring function is a linear transformation function or a nonlinear transformation function; based on preset weights, perform a weighted summation of the score values ​​of each dimension to obtain a weighted summation result; the weights satisfy the condition that the value range is between 0 and 1 and the sum of all weights is 1; calculate the real-time network status score based on the weighted summation result.

[0049] In one embodiment, customer information includes average revenue per customer, customer online time, customer contract status, and customer historical complaint data; the calculation module 23 can be configured to: use average revenue per customer, customer online time, customer contract status, and customer historical complaint data as hyperparameters of a preset model; input customer information and network fault information into the preset model to predict customer churn risk value; and back-map the customer churn risk value to a customer value score.

[0050] In one embodiment, the preset model is trained using at least one of random forest, logistic regression, and XGBoost algorithms.

[0051] In one embodiment, the response module 22 can be configured to: preprocess customer complaint information and input it into a Chinese text analysis model; the Chinese text analysis model is obtained by adjusting the complaint data containing fault labels based on a pre-trained Chinese language model; process the customer complaint information through the Chinese text analysis model and output at least one classification label for indicating the fault domain; determine the management department corresponding to the fault domain classification label according to a predefined mapping relationship, thereby generating network fault information containing the fault domain and the corresponding management department.

[0052] In one embodiment, the scoring module 24 can be configured to: calculate a comprehensive score of the routing result based on the urgency score of the complaint, the real-time network status score, the customer value score, and preset weight coefficients; wherein the weight coefficients include at least a first weight corresponding to the urgency score, a second weight corresponding to the real-time network status score, and a third weight corresponding to the customer value score.

[0053] In one embodiment, the optimization module 25 can be configured to: compare the comprehensive score of the routing result with a preset threshold; when the comprehensive score of the routing result is greater than the preset threshold, determine the management department corresponding to the routing result greater than the preset threshold as the department to be routed; take the management department with the highest comprehensive score among the departments to be routed as the target routing department; the target routing department is used to process customer complaint information.

[0054] In one embodiment, the optimization module 25 can be configured to: determine the routing result as a result to be dispatched when the comprehensive score of the routing result is not greater than a preset threshold; wherein the result to be dispatched is subject to secondary review and dispatch by human customer service.

[0055] This invention provides a customer complaint processing device. The device can be implemented through software, hardware, or a combination of both. From a hardware perspective, in addition to a CPU, memory, network interface, and non-volatile memory, the device in the embodiment typically includes other hardware, such as a forwarding chip responsible for processing messages. Taking software implementation as an example, as a logical device, it is formed by the CPU of the device loading the corresponding computer program instructions from the non-volatile memory into memory for execution.

[0056] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the customer complaint handling method of any of the above embodiments.

[0057] In addition to the methods and devices described above, embodiments of the present invention may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the customer complaint handling methods described above according to various embodiments of the present invention.

[0058] According to another aspect of the present invention, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; and a processor for performing the customer complaint processing method of any of the above embodiments.

[0059] Furthermore, embodiments of the present invention may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the customer complaint handling methods described above according to various embodiments of the present invention.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for handling customer complaints, characterized in that, include: Obtain customer complaint information; Responding to customer complaints, obtaining customer information and network failure information; Based on the customer information and the network fault information, calculate the complaint urgency score, the real-time network status score, and the customer value score. A comprehensive score is calculated based on the complaint urgency score, the network real-time status score, and the customer value score. Network resource allocation is optimized based on the comprehensive score.

2. The customer complaint handling method according to claim 1, characterized in that, The network failure information includes the number of affected customers and the duration of the failure; the customer information includes the customer service level agreement level. The calculation of the urgency score for a complaint includes: The number of affected customers is calculated by multiplying the first weight, the customer service level agreement level by the second weight, and the duration of the outage by the third weight. The urgency score of a complaint is calculated by adding the product of the number of affected customers multiplied by the first weight, the product of the customer service level agreement level multiplied by the second weight, and the product of the duration of the fault multiplied by the third weight.

3. The customer complaint handling method according to claim 1, characterized in that, The network fault information includes the number of alarms and the alarm level; The calculation of the network real-time status score includes: Obtain the number and level of alarms across multiple dimensions; where each dimension corresponds to at least one alarm level and alarm count. According to the preset scoring function, the alarm level and alarm quantity of each dimension are processed to obtain the score value of each dimension; the scoring function is a linear transformation function or a nonlinear transformation function. Based on preset weights, the score values ​​of each dimension are summed in a weighted manner to obtain a weighted sum result; the weights satisfy the condition that the value ranges from 0 to 1 and the sum of all weights is 1; The real-time network status score is calculated based on the weighted summation result.

4. The customer complaint handling method according to claim 1, characterized in that, The customer information includes average revenue per customer, customer online time, customer contract status, and customer historical complaint data. Calculating the customer value score includes: The average revenue per customer, the customer's online time, the customer's contract status, and the customer's historical complaint data are used as hyperparameters of the preset model. Customer information and network failure information are input into the preset model to predict the risk value of customer churn; The customer churn risk value is then mapped inversely to a customer value score.

5. The customer complaint handling method according to claim 4, characterized in that, The preset model is trained using at least one of the following algorithms: random forest, logistic regression, and XGBoost.

6. The customer complaint handling method according to claim 1, characterized in that, In response to customer complaints, obtain customer information and network failure information, including: The customer complaint information is preprocessed and then input into a Chinese text analysis model; the Chinese text analysis model is obtained by adjusting the complaint data containing fault labels based on a pre-trained Chinese language model. The customer complaint information is processed by the Chinese text analysis model, and at least one classification label is output to indicate the fault domain. Based on the predefined mapping relationship, the management department corresponding to the fault domain classification label is determined, thereby generating network fault information containing the fault domain and the corresponding management department.

7. The customer complaint handling method according to claim 6, characterized in that, A comprehensive score is calculated based on the complaint urgency score, the real-time network status score, and the customer value score, including: The comprehensive score of the routing result is calculated based on the complaint urgency score, the network real-time status score, the customer value score, and preset weight coefficients; wherein the weight coefficients include at least a first weight corresponding to the urgency score, a second weight corresponding to the network real-time status score, and a third weight corresponding to the customer value score.

8. The customer complaint handling method according to claim 7, characterized in that, Optimize network resource allocation based on the comprehensive score, including: The overall score of the routing results is compared with a preset threshold; When the overall score of the routing result is greater than a preset threshold, the management department corresponding to the routing result that is greater than the preset threshold is determined as the department to be routed; The management department with the highest overall score among the departments to be routed is designated as the target routing department; the target routing department is used to process the customer complaint information.

9. The customer complaint handling method according to claim 1, characterized in that, Optimizing network resource allocation based on the comprehensive score also includes: When the overall score of the routing result is not greater than the preset threshold, the routing result is determined to be a result to be dispatched; wherein, the result to be dispatched is subject to secondary review and dispatch by human customer service.

10. A customer complaint handling device, characterized in that, include: The acquisition module is used to acquire customer complaint information; The response module is used to respond to customer complaints and obtain customer information and network failure information. The calculation module is used to calculate the complaint urgency score, the real-time network status score, and the customer value score based on the customer information and the network fault information. The scoring module is used to calculate a comprehensive score based on the complaint urgency score, the network real-time status score, and the customer value score. The optimization module is used to optimize network resource allocation based on the comprehensive score.

Images