Complaint work order quality evaluation method, device, equipment, medium and program product

Abstract

The invention discloses a complaint work order quality evaluation method, device and equipment, a medium and a program product, and relates to the technical field of mobile communication, the method comprises the following steps: firstly, obtaining a complaint work order evaluation type of a communication service, and determining a prior probability according to a statistical value of the complaint work order evaluation type of historical complaint work order data; the class conditional probability is determined according to the statistical value of the target business feature of the historical complaint work order data; wherein the target business feature is a business feature in which an expected value of a variance expansion factor in the historical complaint work order data is greater than a set threshold value; and obtaining to-be-evaluated complaint work order data, determining the posterior probability of the to-be-evaluated complaint work order according to the prior probability, the class condition probability and the to-be-evaluated complaint work order data, and determining the quality classification of the to-be-evaluated complaint work order according to the posterior probability. According to the embodiment of the invention, the quality evaluation efficiency of the complaint work order is improved.

Description

Technical Field

[0001] This application belongs to the field of mobile communication technology, and in particular relates to a method, apparatus, equipment, medium and program product for evaluating the quality of complaint work orders. Background Technology

[0002] With the development of mobile communication services, from meeting users' daily needs to improving user perception and service satisfaction, the construction of mobile communication complaint channels and platforms has also become more comprehensive.

[0003] In daily complaint handling, there is a wide variety of complaint content. Current technologies for quality assessment of complaint content often involve text analysis of abnormal complaints, using keyword clustering, manual sampling and labeling of abnormal complaints, and then determining whether a complaint belongs to an abnormal cluster based on the proportion of abnormal complaints within that cluster. In other words, current technologies can only identify whether a complaint is abnormal after processing. Therefore, the efficiency of current technologies for quality assessment of complaint work orders is low. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, medium, and program product for quality assessment of complaint work orders, in order to solve the problem of low efficiency in the quality assessment of complaint work orders in the prior art.

[0005] In a first aspect, embodiments of this application provide a method for evaluating the quality of complaint work orders, the method comprising: The system obtains the evaluation type of complaint tickets for communication services, determines the prior probability based on the statistical value of the evaluation type of complaint ticket data from historical complaint ticket data, and determines the class-conditional probability based on the statistical value of the target service feature from historical complaint ticket data; wherein, the target service feature is a service feature in which the expected value of the variance inflation factor in historical complaint ticket data is greater than a set threshold. Obtain the complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

[0006] Secondly, embodiments of this application provide an apparatus for evaluating the quality of complaint work orders, the apparatus comprising: The acquisition module is used to acquire the evaluation type of the complaint work order for communication services, the prior probability determined based on the statistical value of the evaluation type of the complaint work order based on the historical complaint work order data, and the class conditional probability determined based on the statistical value of the target service feature based on the historical complaint work order data; wherein, the target service feature is the service feature in which the expected value of the variance inflation factor in the historical complaint work order data is greater than a set threshold. The determination module is used to acquire complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

[0007] Thirdly, embodiments of this application provide a terminal device, the device including: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the method for quality assessment of complaint work orders as described in the first aspect.

[0008] Fourthly, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the complaint work order quality assessment method as described in the first aspect.

[0009] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform a method for evaluating the quality of complaint work orders as described in the first aspect.

[0010] This application provides a method, apparatus, device, medium, and program product for evaluating the quality of complaint work orders. The method obtains the evaluation type of a complaint work order for a communication service, determines a prior probability based on statistical values ​​of the evaluation types of complaint work order data, and determines a class-conditional probability based on statistical values ​​of target service features from historical complaint work order data. The target service feature is a service feature where the expected value of the variance inflation factor in historical complaint work order data is greater than a set threshold. By statistically analyzing the prior probability from historical complaint work order data, combining it with the variance inflation factor to screen target service features and calculate class-conditional probabilities, the objectivity and reliability of the probability data are ensured. Furthermore, by eliminating redundant features, information interference is reduced, and the distinguishability of features for evaluation types is improved, laying a high-quality data foundation for subsequent classification. The process involves acquiring complaint work order data to be evaluated, determining the posterior probability of each complaint work order based on its prior probability, class-conditional probability, and the data, and then determining its quality classification based on the posterior probability. By fusing the prior probability, class-conditional probability, and actual characteristics of the complaint work order to calculate the posterior probability, a precise correlation between historical statistical patterns and individual work order characteristics is achieved, and the posterior probability can quantify the classification confidence level. Therefore, this embodiment improves the efficiency of complaint work order quality evaluation. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating the method for evaluating the quality of complaint work orders provided in this application embodiment; Figure 2 This is a flowchart illustrating the method for obtaining prior probability and class conditional probability through statistics provided in the embodiments of this application; Figure 3 This is a flowchart illustrating the method for determining the expected value of the variance inflation factor provided in the embodiments of this application. Figure 4 This is a flowchart illustrating the method for determining prior probabilities and class conditional probabilities provided in an embodiment of this application. Figure 5 This is a schematic diagram of the device for evaluating the quality of complaint work orders provided in the embodiments of this application; Figure 6 This is a schematic diagram of the structure of the terminal device provided in the embodiments of this application. Detailed Implementation

[0013] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0014] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0015] With the development of mobile communication services, from meeting users' daily needs to improving user experience and satisfaction, the construction of mobile communication complaint channels and platforms has become more comprehensive. The complaint handling support system integrates various network complaint handling IT support systems, constructs a unified network complaint support and handling platform, provides centralized, standardized, and intelligent complaint handling and management support methods, improves the front-end customer service resolution capabilities, reduces complaint ticket processing time, shortens the average complaint processing time, promotes network quality improvement, and enhances productivity through machine learning algorithms. It supports complaints from various departments across the network, including marketing, customer service front and back offices, monitoring, and various professional departments. Daily complaint handling involves a wide variety of complaints. With a massive volume of complaints, it is difficult to efficiently assess the reasonableness of complaints manually, requiring significant manpower and resources for processing. When user complaints are untrue or involve repeated false complaints, they are only categorized as abnormal complaints afterward, making it impossible to predict normal complaints in advance. Processing all complaints as normal complaints leads to a significant waste of resources in daily operations.

[0016] Existing solutions involve text analysis of abnormal complaints, using keyword clustering, manual sampling and labeling of abnormal complaints, and then confirming whether a complaint belongs to an abnormal cluster based on the proportion of abnormal complaints within that cluster. These existing solutions only categorize and output complaints after the fact, failing to pre-assess users with abnormal complaints. The lack of pre-judgment tools for abnormal complaint handling forces the process to rely solely on post-event identification and manual labeling, resulting in wasted repetitive labor. Furthermore, subjective manual labeling is susceptible to human error; the skill level of the personnel handling the complaints can influence the results, making it impossible to objectively and quantitatively ensure data accuracy.

[0017] To address the problems of existing technologies, embodiments of this application provide a method, apparatus, device, medium, and program product for evaluating the quality of complaint work orders. The method acquires the evaluation type of a complaint work order for a communication service, determines a prior probability based on statistical values ​​of the evaluation types of complaint work order data from historical complaint work order data, and determines a class-conditional probability based on statistical values ​​of target service features from historical complaint work order data. The target service feature is a service feature where the expected value of the variance inflation factor in historical complaint work order data is greater than a set threshold. By statistically analyzing the prior probability based on historical complaint work order data, and combining it with the variance inflation factor to screen target service features and calculate class-conditional probabilities, the objectivity and reliability of the probability data are ensured. Furthermore, by eliminating redundant features, information interference is reduced, and the distinguishability of features for evaluation types is improved, laying a high-quality data foundation for subsequent classification. The process involves acquiring complaint work order data to be evaluated, determining the posterior probability of each complaint work order based on its prior probability, class-conditional probability, and the data, and then determining its quality classification based on the posterior probability. By fusing the prior probability, class-conditional probability, and actual characteristics of the complaint work order to calculate the posterior probability, a precise correlation between historical statistical patterns and individual work order characteristics is achieved, and the posterior probability can quantify the classification confidence level. Therefore, this embodiment improves the efficiency of complaint work order quality evaluation.

[0018] The following section first introduces the method for evaluating the quality of complaint work orders provided in the embodiments of this application.

[0019] Figure 1 A flowchart illustrating a method for evaluating the quality of complaint work orders according to an embodiment of this application is shown. Figure 1 As shown, the method may include: S101 and S102.

[0020] S101, obtain the complaint ticket evaluation type of the communication service, the prior probability determined based on the statistical value of the complaint ticket evaluation type in the historical complaint ticket data, and the class conditional probability determined based on the statistical value of the target service feature in the historical complaint ticket data; wherein, the target service feature is a service feature in the historical complaint ticket data whose expected value of variance inflation factor is greater than a set threshold.

[0021] The complaint work order assessment type is a category categorized according to preset standards, used to define the quality of the work order or the severity of the problem. Prior probability is the initial probability of a certain assessment type occurring based on historical data statistics, before considering the specific characteristics of the work order to be assessed; it reflects the inherent distribution of that type in historical data. Class-conditional probability is the probability that a work order possesses a certain target business characteristic given that it belongs to a certain assessment type; it describes the degree of correlation between the characteristic and the assessment type. Variance inflation factor (VIF) is an indicator that measures the degree of multicollinearity among multiple business characteristics; the larger the VIF value, the stronger the linear correlation and the more information redundancy among the characteristics. Expected value is the statistical average of the VIF values ​​of that business characteristic in historical data. Target business characteristic is a highly discriminative and effective feature selected from all business characteristics, with an expected VIF value greater than a set threshold.

[0022] The prior probabilities in this application are obtained based on objective statistics of historical data, providing an initial probability basis for subsequent posterior probability calculations and avoiding subjective assumptions. Class-conditional probabilities establish the association between target features and evaluation types. Furthermore, by using the expected value of the variance inflation factor, features with strong linear correlation and redundant information are eliminated, reducing subsequent computation while improving the distinguishability of features for evaluation types. This avoids the problem of class-conditional probability calculation bias caused by redundant features, which reduces classification accuracy.

[0023] S102, obtain the complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

[0024] Among them, the complaint work order data to be evaluated refers to newly added complaint work order information that has not yet been classified into a quality category, and is the target object of the evaluation. The posterior probability is the probability of each evaluation type occurring after obtaining the specific target business characteristics of the work order to be evaluated, combined with the prior probability and class-conditional probability; it reflects the confidence level in judging that the work order belongs to a certain category based on its actual characteristics. Quality classification is the final result of classifying the work order to be evaluated into the corresponding evaluation type based on the magnitude of the posterior probability.

[0025] In some embodiments, feature values ​​of the target business in the complaint work order data to be evaluated can be extracted. Based on the feature values, prior probabilities, and class-conditional probabilities of the target business features, the posterior probabilities of the complaint work order to be evaluated belonging to different quality categories can be calculated. Finally, the posterior probabilities of different quality categories are compared, and the quality category with the highest posterior probability is taken as the quality category of the complaint work order to be evaluated.

[0026] The prior probability in this application only reflects the historical type distribution and cannot reflect the individual differences of the work order to be evaluated. By combining the actual characteristics of the work order to be evaluated with the class conditional probability, personalized probability correction can be achieved, making the classification result more in line with the actual situation of the work order. At the same time, the posterior probability quantifies the confidence that the work order belongs to a certain class. The clear classification rules can ensure the consistency of the classification results. Compared with rule judgment based on features, it is more objective and avoids the subjective differences of human judgment.

[0027] In summary, this application's embodiments employ a two-step collaborative approach: obtaining prior probabilities and class-conditional probabilities, and calculating posterior probabilities for quality classification. First, prior probabilities are statistically analyzed based on historical complaint work order data. Then, target business features are screened using variance inflation factors, and class-conditional probabilities are calculated. This ensures the objectivity and reliability of the probability data while reducing information interference by eliminating redundant features, thus improving the distinguishability of features for assessment types and laying a high-quality data foundation for subsequent classification. Second, by fusing prior probabilities, class-conditional probabilities, and the actual characteristics of the work order to be evaluated to calculate posterior probabilities, a precise correlation is achieved between historical statistical patterns and individual work order characteristics. The posterior probability can quantify classification confidence. Therefore, this application's embodiments improve the efficiency of complaint work order quality assessment.

[0028] In some embodiments, such as Figure 2 As shown, before obtaining the prior probability of the complaint ticket evaluation type of the communication service determined by the statistical value of the complaint ticket evaluation type based on the historical complaint ticket data, and the class conditional probability determined by the statistical value of the target service characteristics based on the historical complaint ticket data, the method may further include: S201 to S204.

[0029] S201, Obtain multiple historical complaint work order data and feature information of historical complaint work orders for communication services. The feature information includes multiple service features and preset correlation data of service features.

[0030] The historical complaint work order data consists of past complaint records that have already undergone quality classification, used for model training. Business features are key attributes that differentiate complaint quality. Predefined correlation data are quantitative data on the correlations between predefined business features.

[0031] In some embodiments, the data collection scope is first defined according to the communication service, and three types of core data are extracted from the communication service system, including newly added real-time data, i.e. work orders to be evaluated, existing data with classification labels, i.e. historical work orders, and matrix data sorted out based on past complaint handling experience, i.e. feature correlation data, and ensure that the data fields are complete and the format is consistent.

[0032] The data sources in this application focus on communication service scenarios and are tailored to actual complaint handling needs; the relevant data are based on experience set from historical data to ensure the rationality of feature associations; the three types of data support each other and provide comprehensive and reliable basic support for subsequent steps.

[0033] In some embodiments, when acquiring complaint work order data to be evaluated, historical complaint work order data, and feature information of historical complaint work orders for communication services, the acquired information is preprocessed, which may include operations such as removing duplicate work orders and invalid work orders, and correcting feature value errors.

[0034] This application embodiment reduces the impact of noisy data on subsequent evaluations by preprocessing the acquired information, thereby improving the relevance and accuracy of the evaluation.

[0035] S202, determine the weight of each business feature based on the preset correlation data of multiple business features, and determine the expected value of the variance inflation factor of each business feature based on historical complaint work order data, business features and weights.

[0036] The weight of a business feature is a numerical value that measures the importance of the feature in the evaluation. The variance inflation factor is an indicator that measures multicollinearity among business features, and its expected value is the average of the variance inflation factors of multiple data points.

[0037] In some embodiments, firstly, weights are assigned to each business feature based on preset correlation data; based on overall historical complaint work order data, regression analysis is performed with any business feature as the dependent variable and other business features as independent variables to obtain the determination coefficient of each business feature, and the expected value of the variance inflation factor of each business feature is determined according to the determination coefficient and weight of each business feature.

[0038] This application's embodiments improve the accuracy of subsequent model evaluation by determining the weights of business features and highlighting those with a greater impact on complaint quality. The expected value of the variance inflation factor is calculated using overall historical data, ensuring both sufficient data samples and the ability to effectively screen out features without multicollinearity, thus avoiding feature redundancy that could lead to complex model calculations and biased results.

[0039] S203, identify business features whose expected value is greater than a set threshold as target business features, and generate business feature templates based on target business features.

[0040] The threshold is set as a critical value for determining whether a feature is a target feature. The target business feature is a business feature whose expected variance inflation factor is greater than the set threshold. The business feature template is a standardized data structure containing the target business feature and core identifier fields, used to unify the data input format.

[0041] In some embodiments, the expected value of the variance inflation factor of each business feature is compared with a set threshold, and business features with expected values ​​greater than the set threshold are selected as target business features, i.e., strongly correlated business features with expected values ​​not exceeding the set threshold are eliminated. Then, the core identifier fields of the target business features are integrated to generate a structured business feature template.

[0042] This application's embodiments filter features with expected values ​​greater than a set threshold as target business features, eliminating strongly correlated features to avoid feature redundancy that leads to model computational complexity and decreased accuracy. Standardized templates unify the format of historical data and data to be evaluated, ensuring data consistency.

[0043] S204. Generate a prior dataset based on the business feature template and historical complaint work order data, and statistically analyze the data in the prior dataset to obtain the prior probability and class conditional probability of the complaint work order evaluation type.

[0044] The prior dataset is a labeled historical dataset organized based on business feature templates. The prior probability is the probability that a complaint order belongs to a certain assessment type without considering features; that is, the proportion of samples of that type in the total prior dataset. The class-conditional probability is the probability of a certain business feature appearing given an assessment type.

[0045] In some embodiments, fields are extracted from historical work orders according to business feature templates, and evaluation type labels are added to generate a prior dataset. The proportion of samples in different quality categories in the prior dataset is counted, and the prior probability reflecting the overall distribution of different categories is calculated based on the proportion of samples in different quality categories. The frequency of each business feature in different quality categories is counted, and the class conditional probability used to quantify the association strength between features and categories is calculated based on the frequency of each business feature in different quality categories.

[0046] The prior dataset in this application is based on a standardized template, with a unified data structure, which improves the efficiency of model training; the prior probability provides an initial probability reference, and the class conditional probability quantifies the correlation between features and types; the probability statistics are based on objective data throughout the process without human intervention, ensuring the objectivity of the results.

[0047] This application embodiment calculates weights using pre-set relevance data to ensure the objectivity of feature importance assessment. It combines historical work order data to calculate the expected value of the variance inflation factor, ensuring the completeness of feature independence assessment and improving the accuracy of feature selection. Features are selected by comparing the expected value with a threshold, generating a business feature template. Strongly correlated or unquantifiable redundant business features are eliminated, reducing the complexity of subsequent model calculations. Target feature values ​​and actual categories are directly extracted from historical work order data using the business feature template, allowing for rapid assembly into a standardized prior dataset. Prior probabilities and class-conditional probabilities are obtained statistically, eliminating the need for complex calculations and improving processing efficiency. The posterior probability of the complaint work order to be evaluated is determined based on the prior probability, class-conditional probability, and the complaint work order data to be evaluated. The quality classification of the complaint work order to be evaluated is then determined based on the posterior probability. The prior probability and class-conditional probability statistically derived from historical data are used to evaluate the work order to be evaluated, shortening the time required for real-time evaluation.

[0048] In some embodiments, by accumulating extensive experience in handling complaints, the mobile communication user service characteristics of abnormal and normal complaint tickets are analyzed. These characteristics must be representative, independent, and unrelated. Relevant data from complaint tickets is compiled, and combined with complaint handling experience, scenario-based business characteristics are mined from the complaint tickets, outputting a business feature matrix. First, a correlation analysis is performed on the business feature matrix. By analyzing the degree of correlation between feature variables, the intrinsic relationships between multiple feature variables are analyzed. By observing the correlation matrix, the interaction and group relationships between feature variables are discovered, and business feature weighting factors are designed to provide objective weighting factors for subsequent calculations, enhancing algorithm performance. The preset correlation data for business features is shown in Table 1, which includes fields for normal indicators, fixed location, repeated complaints, frequently switched-off devices, common complaints, contradictory content, false content, outrageous demands, malicious content, and ambiguous addresses. Y indicates that two fields are correlated, and N indicates that they are not correlated.

[0049] Table 1. Related Tables of Business Feature Matrix In some embodiments, determining the weight of each business feature based on preset correlation data of multiple business features may include: Based on preset correlation data of multiple business features, the number of business features that are not related in business is counted. The first business feature is any one of the multiple business features, and the second business feature is any business feature other than the first business feature. The number of business features that are not related in business is the number of features that are determined to be unrelated according to the preset correlation data among the first business feature and all second business features. Calculate the ratio of the number of business-irrelevant first business features and second business features to the total number of business features, and determine the weight of the first business feature; where the total number of business features is the sum of the first business features and the second business features.

[0050] This application's embodiments determine weights by statistically analyzing the number of irrelevant first and second business features and calculating their ratios, effectively highlighting the importance of business features; it is based entirely on objective statistics of preset relevance data, avoiding subjective human intervention, and the results are authentic, reliable, and reproducible; the calculation logic is simple, with low complexity, and is suitable for rapid computation in scenarios with large amounts of data.

[0051] In some embodiments, the formula for calculating the weight of a business feature can be: in, As the weighting factor, based on the relevant table of the business feature matrix, the count of Y is expressed as: Counting N is represented as We use the proportion of business-irrelevant factors N to generate weighting factors, thereby enhancing the independence and importance of business features.

[0052] In some embodiments, such as Figure 3 As shown, the expected value of the variance inflation factor for each business feature is determined based on historical complaint work order data, business characteristics, and weights, and may include: S301 to S304.

[0053] S301 splits historical complaint work order data into multiple data subsets.

[0054] The data subsets are several sub-data sets obtained by splitting the dataset, which are used to reduce the bias of a single data sample.

[0055] In some embodiments, historical complaint work order data can be randomly split into multiple data subsets, with historical complaint work orders as the smallest unit.

[0056] This application embodiment ensures that the selected data has the same probability by randomly splitting historical complaint work order data, eliminating the influence of human preference and ensuring the objectivity of data subset splitting.

[0057] In some embodiments, complaints can be categorized according to historical complaint work order data, and then random sampling can be performed on the historical complaint work orders corresponding to each category of complaint work order data to ensure that the proportion of historical complaint work order data for each category of each subset is the same.

[0058] The embodiments of this application solve the problem of imbalance in the proportion of classification types that may be caused by random sampling by using stratified sampling, thereby improving the representativeness of data of different classification types.

[0059] S302, for each data subset, perform regression analysis with the second business feature as the independent variable and the first business feature as the dependent variable to obtain the determination coefficient of the first business feature; the first business feature is any one of the multiple business features, and the second business feature is any business feature other than the first business feature.

[0060] Among them, regression analysis is used to quantify the linear explanatory power of independent variables on dependent variables by constructing a model; the coefficient of determination is an indicator that measures the linear explanatory power of the second business characteristic on the first business characteristic.

[0061] In some embodiments, the coefficient of determination can take values ​​between 0 and 1, with the closer to 1 indicating stronger explanatory power.

[0062] In some embodiments, for each data subset, each business feature is selected as the first business feature, the remaining business features are selected as the second business features, a linear regression model of the first business feature and the second business feature is constructed, the model parameters are fitted by the least squares method, and finally the determination coefficient of the first business feature under the subset is calculated.

[0063] The embodiments of this application can objectively quantify the influence of the second business feature on the first business feature through regression analysis. Furthermore, the linear regression model is simple and easy to understand, has low computational complexity, and is suitable for rapid processing of big data subsets. The determination coefficient results are quantifiable, which facilitates consistency comparison between different subsets. It can effectively screen out the linear correlation between features, providing an objective basis for subsequent judgment of the independence of business features.

[0064] S303, determine the variance inflation factor of the first business feature corresponding to each data subset based on the determination coefficient of the first business feature and the weight of the business feature.

[0065] In some embodiments, the formula for calculating the variance inflation factor can be: in, Features variance inflation factor For other features This feature The determination coefficients obtained from regression analysis are further enhanced with business weight factors to improve their business specificity. A higher value indicates a strong linear correlation between the feature and other features.

[0066] S304. Calculate the mean of multiple variance inflation factors for all data subsets of each first business feature to obtain the expected value of the variance inflation factor of the first business feature.

[0067] In some embodiments, this application calculates the variance inflation factor of each feature based on the data flywheel methodology. This method iterates through the dataset of business features using both existing and incremental data to ensure the completeness of the business features. The existing and incremental data are split into n random data subsets, and the variance inflation factor of each subset is calculated separately. After multiple rounds of data iteration, the expected variance inflation factor of each feature's performance is finally obtained.

[0068] This application's embodiments avoid the bias of single data by splitting data, quantify the linear correlation between features through regression analysis, and further improve the robustness of the results by solving the mean value of the variance inflation factor. The entire process is based on objective data calculation without human subjective intervention, which improves the objectivity of subsequent target business feature determination.

[0069] In some embodiments, the expected value of the variance inflation factor can be calculated using the following formula: in, Features The expected variance inflation factor, such as A value greater than 5 indicates a strong correlation between the features, which is marked as Y; otherwise, it is marked as N.

[0070] In some embodiments, the expected value of the variance inflation factor of the business features is calculated to obtain the complaint business feature performance table as shown in Table 2, where Y indicates that the business features have a strong correlation and N indicates that the business features do not have a strong correlation.

[0071] Table 2. Performance Characteristics of Complaint Handling Services In some embodiments, the present application does not use manual feature annotation. Feature fields such as "normal indicators", "fixed location" (referring to fixed user location information) and "frequently turned off" can be obtained through existing call detail record statistics. "Repeated complaints" (more than 2 complaints within 1 month) and "common complaints" (2 complaints with the same problem within 1 month) can be obtained through statistics of existing complaint work orders.

[0072] In some embodiments, business characteristics with expected values ​​greater than a set threshold are identified as target business characteristics. Based on the target business characteristics, the business characteristic library table shown in Table 3 is obtained, where Y represents the determined value, namely, normal indicators, fixed location, repeated complaints, frequent shutdowns, and common complaints, and N represents the negative value, namely, non-normal indicators, non-fixed location, non-repeated complaints, non-frequent shutdowns, and non-common complaints.

[0073] Table 3 Business Feature Library In some embodiments, finally, a feature template is generated by combining the key fields of the feature library. Only the "complaint work order number" (work order identifier used to evaluate abnormal complaints), "complaint issue" (records of common complaints), "normal indicators", "fixed location", "repeated complaints", "frequent shutdowns", and "common complaints" are retained. This results in the business feature template table shown in Table 4. Based on the generated feature template, the corresponding preprocessed data template is output. Y and N are represented by binary 1 and 0 respectively, which can improve the computational efficiency when performing matrix operations, such as transpose operations and distributed computing.

[0074] Table 4 Business Feature Template Table In some embodiments, regression analysis is performed on each data subset, with the second business feature as the independent variable and the first business feature as the dependent variable, to obtain the determination coefficient of the first business feature, which may include: Establish a linear relationship model with the second business characteristic as the independent variable and the first business characteristic as the dependent variable; wherein, the linear relationship model is a mathematical expression describing the linear association between the independent and dependent variables, used to quantify the linear influence relationship between the characteristics.

[0075] The coefficients of each independent variable in the linear model are determined based on the data in the current subset of data, so that the difference between the actual value of the dependent variable and the predicted value obtained based on the linear model is less than a set threshold. The coefficients of each independent variable are the weight parameters of each independent variable in the linear model. The set threshold is a critical value that measures the acceptable degree of difference between the actual value and the predicted value of the dependent variable. Based on the coefficients of each independent variable and the linear relationship model, the predicted value is calculated, and the coefficient of determination is calculated according to the degree of difference between the actual value and the predicted value of the dependent variable. The predicted value is the estimated value of the dependent variable calculated by substituting the actual value of the independent variable into the linear model with the determined coefficients.

[0076] This application's embodiments construct an objective quantitative analysis method adapted to the characteristics of communication complaint work orders by establishing a linear model, solving for the coefficients of independent variables, and calculating the coefficients of determination. The linear model adapts to the feature data format, the threshold constraint of the difference value improves the fitting accuracy, and the calculated coefficients of determination quantify the strength of the linear correlation between features, providing a reliable input for the subsequent calculation of the expected value of the variance inflation factor.

[0077] In some embodiments, a priori dataset is generated based on business feature templates and historical complaint work order data. A training dataset for the preprocessed data template can be collected. The feature set corresponding to the training samples is The category set is .

[0078] In some embodiments, the training dataset table is shown in Table 5. Each sample includes the following fields: training dataset sample number (n), normal metrics, fixed location, repeated complaints, frequent shutdowns, common complaints, and abnormal complaints (C1=Y, C2=N). Therefore, the feature set is... That is, F = {normal indicators, fixed location, repeated complaints, frequent shutdowns, common complaints}; and the category set is... ,Right now .

[0079] Table 5 Training Dataset Table In some embodiments, such as Figure 4 As shown, the prior probability and class conditional probability of the complaint work order evaluation type are obtained from the data in the statistical prior dataset, which may include: S401 to S403.

[0080] S401, respectively count the ratio of the number of complaint work orders of different evaluation types in the prior dataset to the total number of complaint work orders, and obtain the initial prior probability of different complaint work order evaluation types.

[0081] The initial prior probability is the ratio of the number of samples of a certain evaluation type to the total number of samples in the prior dataset, reflecting the basic distribution of that type in historical data.

[0082] In some embodiments, data can be derived from prior data. and The prior probability.

[0083] In some embodiments, the prior probability table for the quality assessment of complaint work orders is shown in Table 6, where C1 represents the abnormal complaint category and C2 represents the normal complaint category.

[0084] The initial prior probability in this application is the basis for subsequent probability calculations, directly reflecting the natural distribution of the two types of complaints in historical data, and providing an initial reference for posterior probability calculations.

[0085] S402, determine the prior probability of different complaint work order evaluation types based on the initial prior probability of the complaint work order evaluation type and the preset weight of the complaint work order evaluation type.

[0086] The prior probability is the weighted sum of the initial prior probability and the corresponding preset weight.

[0087] In some embodiments, an adaptive factor is used to enhance the prior probability learning of Naive Bayes: The prior probability calculation based on the training dataset in this application is limited by the size of the training data, affecting the accuracy of the prior probability. This proposal introduces a preset weight for the complaint ticket evaluation type. The prior probabilities are reinforced through learning, combining the prior probabilities from existing training data with adaptive coefficients derived from real-time feedback and continuous data acquisition during the production process. And adjust the prior probability dynamically in real time.

[0088] in, The quantity of each category for work order quality assessment. This represents the total number of work orders for the day. For the preset weights of the evaluation type, when the work order quality evaluation category count is not 0, the adaptive coefficient updates the prior probability weight in real time; when the work order quality evaluation category count is 0, the adaptive coefficient is 1 and does not need to be updated (that is, when the evaluation category does not change, there is no need for self-learning to adjust in real time).

[0089] The initial prior probability in this application embodiment only reflects the distribution of historical data and may not take into account actual business needs; the preset weight can adjust the type priority, so that the final prior probability is more in line with business goals, thereby flexibly adapting to business needs.

[0090] S403, respectively calculate the first statistical value of the occurrence of business features in complaint work orders of different evaluation types in the prior dataset, and the second statistical value of the number of times the business features occur in complaint work orders of the same evaluation type, and determine the class conditional probability based on the first statistical value and the second statistical value.

[0091] In some embodiments, class-conditional probabilities may include class-conditional probabilities of anomalous complaints. The conditional probability of a normal complaint .

[0092] In some embodiments, the formula for calculating class conditional probability can be: in, For a given category Features The conditional probability, For this feature exist The number of times the category appears For the category The sum of the total number of occurrences of all features, This serves as an identifier for a specific piece of data.

[0093] The initial prior probability in this application embodiment is based on the full data statistics of the prior dataset, ensuring the authenticity of the basic distribution; the preset weight adjustment adapts to the business priority, solving the problem of the disconnect between the pure data distribution and the business objectives; it provides high-quality input for the subsequent calculation of the posterior probability of the improved Naive Bayes algorithm, filling the gap in the existing technology in the adaptation of probability statistics to business needs.

[0094] In one example, given that the categories include abnormal complaints and normal complaint records, statistics are collected separately for user characteristics Y (yes) and N (no), and then the results are calculated. and The class-conditional probabilities are shown in Table 7, which is a table of conditional probabilities for features classified as abnormal complaints.

[0095] Table 7 Conditional Probability Table of Abnormal Complaint Characteristics In some embodiments, since it is necessary to calculate the joint probability distribution, when the class conditional probability is... However, due to limitations in the size of the training dataset, the calculation results may be subjective. By introducing Laplace smoothing, the extreme case where a small prior dataset leads to a probability of 0 can be eliminated. The Laplace smoothing formula can be: in, For a given category Features The conditional probability, For this feature exist The number of times the category appears For the category The sum of the total number of occurrences of all features, For the identification of a certain data, The specified coefficient is usually 1. The number of statistical features in the prior dataset (a total of 5 features, i.e.) (5).

[0096] In one example, the smoothed abnormal complaint feature class conditional probability table obtained by substituting into the Laplace smoothing formula is shown in Table 8.

[0097] Table 8. Smoothed Conditional Probability Table of Abnormal Complaint Feature Classes In one example, the smoothed conditional probability table of normal complaint feature classes obtained by substituting into the Laplace smoothing formula is shown in Table 9.

[0098] Table 9. Conditional Probability Table of Normal Complaint Characteristics In some embodiments, determining the posterior probability of a complaint ticket to be evaluated based on prior probability, class conditional probability, and complaint ticket data to be evaluated may include: Extract feature values ​​for each target business characteristic from the complaint work order data to be evaluated; where the target business characteristic is the core indicator for evaluating the quality of the complaint work order; and the feature value is the specific value of the target business characteristic. For each complaint work order assessment type, the class conditional probability corresponding to the feature value is determined, and the class conditional probabilities corresponding to all target business features are multiplied together to obtain the joint conditional probability under that type. The joint conditional probability is the probability that all target business feature values ​​of the work order to be assessed will appear simultaneously when a certain complaint assessment type is given, and is obtained by multiplying the class conditional probabilities corresponding to each feature value.

[0099] Multiplying the joint conditional probability by the prior probability of the complaint work order type yields the posterior probability of the complaint work order to be evaluated belonging to different complaint work order evaluation types. The posterior probability is the probability that the work order to be evaluated belongs to a certain evaluation type after considering the feature value of the work order to be evaluated, which is obtained by multiplying the joint conditional probability of that type by the prior probability.

[0100] This application's embodiments extract feature values ​​from each target business characteristic in the complaint work order data to be evaluated. The extraction logic aligns with the business characteristic template, standardizing the extraction results and meeting the needs of subsequent calculations. The joint conditional probability quantifies the overall correlation strength between the work order characteristics to be evaluated and a certain evaluation type. The calculation logic is simple and intuitive, requiring only basic multiplication operations, and is suitable for parallel processing of batch work orders. The posterior probability integrates the prior probability reflecting historical distribution and the joint conditional probability reflecting current work order characteristics, objectively reflecting the true quality category tendency of the work order.

[0101] In some embodiments, the formula for calculating the posterior probability based on Bayes' theorem can be: In the formula, Given observation features Data is categorized The posterior probability, Refer to a given category Features under the conditions Class conditional probability (learned from prior dataset). Category Prior probabilities (learned from prior datasets). Features The marginal probabilities are also treated as standardized constants and do not require further processing.

[0102] Because of the denominator It is a constant, mainly calculated by the numerator. The posterior probability can then be obtained. Introducing the Naive Bayes algorithm, the significance of the naive data lies in the "feature conditional independence assumption," that is, for a known class, it is assumed that all features are mutually independent, and all features of the sample are equally important to the result. Therefore, the class conditional probabilities can also be multiplied and written as a direct product formula as follows: Where n represents the number of features in the prior dataset, Let be the value of the i-th feature in the prior dataset. Since for all categories, Since these are constants and have the same value, they can be ignored when calculating the posterior probability between different categories under the same feature attribute, i.e., complaint work orders. Quality assessment results The formulas for calculating the posterior probability of different quality assessment categories can be as follows: In some embodiments, to improve computational efficiency, a set of maximum value independent variable points is introduced. The function is used to find the categorical variable behind the maximum posterior probability, and its calculation formula can be: in, Represents the feature data When searching for the maximum value of an expression, return C( The categorization variables of this proposal are: ,Right now If you return Time is The maximum value of the expression, which is the exception complaint; return. Time is The maximum value of the expression, which is the normal complaint.

[0103] This application's embodiments, based on data preprocessing generalization templates corresponding to business feature templates, facilitate efficient computation of prior dataset feature learning and aid in the evaluation of Naive Bayes. In the application scenarios of this proposal, to improve computational efficiency, it can be implemented through open-source code and utilized... The function quickly derives the classification. In summary, the complaint work order with the highest posterior probability is classified as either an abnormal or normal complaint, thus achieving the purpose of assessing the quality of the complaint work order.

[0104] In summary, this application's embodiments analyze and design the business characteristics of complaint work orders, and use an improved variance inflation factor based on the data flywheel methodology to evaluate feature performance, generating business feature templates and corresponding data preprocessing. Secondly, an improved Naive Bayes classification algorithm is proposed, enhancing learning ability with prior probabilities, and calculating the posterior probability of the incremental complaint work order quality real-time evaluation category using the learned prior probabilities and class-conditional probabilities, respectively. Finally, to improve computational efficiency, a nested Naive Bayes algorithm with a maximum independent variable point set function is used to find the work order quality evaluation category behind the maximum posterior probability.

[0105] Furthermore, this application's embodiments employ an improved variance inflation factor based on a data flywheel, specifically, generating business weight factors based on business characteristics to enhance the business specificity of the determination coefficient and improve business characteristic performance. The method involves iterating through the user characteristic dataset using both existing and incremental data to ensure the completeness of the business characteristics. The existing and incremental data are split into n random data subsets, and the characteristic performance of each subset is calculated separately. This process involves multiple rounds of data iteration calculations, and finally, the expected performance of each characteristic is obtained.

[0106] Furthermore, this application employs an improved Naive Bayes classification algorithm to enhance the learning ability of prior probabilities, addressing the limitations of prior datasets for business features. It introduces adaptive coefficients for prior probabilities, learns these adaptive coefficients based on real-time feedback from incremental data of the day, and dynamically adjusts the prior probabilities, accurately addressing the insufficient real-time evaluation capabilities of existing technologies in handling complaint work orders.

[0107] Figure 5 This application illustrates an apparatus 500 for evaluating the quality of complaint work orders, which may include: The acquisition module 501 is used to acquire the complaint ticket evaluation type of the communication service, the prior probability determined based on the statistical value of the complaint ticket evaluation type of the historical complaint ticket data, and the class conditional probability determined based on the statistical value of the target service feature of the historical complaint ticket data; wherein, the target service feature is the service feature in which the expected value of the variance inflation factor of the historical complaint ticket data is greater than a set threshold. The determination module 502 is used to acquire complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

[0108] In some embodiments, the apparatus 500 for evaluating the quality of complaint work orders may further include: The acquisition module 501 is also used to acquire multiple historical complaint work order data and feature information of historical complaint work orders for communication services. The feature information includes multiple service features and preset correlation data of the service features. The determination module 502 is further configured to determine the weight of each business feature based on preset correlation data of multiple business features, and to determine the expected value of the variance inflation factor of each business feature based on historical complaint work order data, business features and weights. A generation module is used to determine the business features whose expected value is greater than a set threshold as target business features, and to generate a business feature template based on the target business features; The statistics module is used to generate a prior dataset based on the business feature template and historical complaint work order data, and to calculate the prior probability and class conditional probability of the complaint work order evaluation type by statistically analyzing the data in the prior dataset.

[0109] In some embodiments, the apparatus 500 for evaluating the quality of complaint work orders may further include: The statistics module is also used to count the number of business features that are not related to the first business feature and the second business feature based on preset correlation data of multiple business features; the first business feature is any one of the multiple business features, and the second business feature is any business feature other than the first business feature. The calculation module is used to calculate the ratio of the number of first business features and second business features that are not related in business to the total number of business features, and to determine the weight of the first business feature based on the ratio.

[0110] In some embodiments, the apparatus 500 for evaluating the quality of complaint work orders may further include: The splitting module is used to split historical complaint work order data into multiple data subsets; The analysis module is used to perform regression analysis on each data subset, with the second business feature as the independent variable and the first business feature as the dependent variable, to obtain the determination coefficient of the first business feature; the first business feature is any one of multiple business features, and the second business feature is any business feature other than the first business feature. The determination module 502 is also used to determine the variance inflation factor of the first business feature corresponding to each data subset based on the determination coefficient of the first business feature and the weight of the business feature. The calculation module is also used to calculate the mean of multiple variance inflation factors of all data subsets of each first business feature, and to obtain the expected value of the variance inflation factor of the first business feature.

[0111] In some embodiments, the apparatus 500 for evaluating the quality of complaint work orders may further include: The module is used to establish a linear relationship model with the second business characteristic as the independent variable and the first business characteristic as the dependent variable. The determination module 502 is also used to determine the coefficients of each independent variable in the linear relationship model based on the data in the current data subset, so that the difference between the actual value of the dependent variable and the predicted value obtained based on the linear relationship model is less than a set threshold. The calculation module is also used to calculate the predicted value based on the coefficients and linear relationship model corresponding to each independent variable, and to calculate the coefficient of determination based on the degree of difference between the actual value and the predicted value of the dependent variable.

[0112] In some embodiments, the statistics module is further used to calculate the ratio of the number of complaint work orders of different evaluation types in the prior dataset to the total number of complaint work orders, so as to obtain the initial prior probability of different complaint work order evaluation types. The determination module 502 is also used to determine the prior probability of different complaint work order evaluation types based on the initial prior probability of the complaint work order evaluation type and the preset weight of the complaint work order evaluation type. The determination module 502 is further used to separately calculate the first statistical value of the occurrence of business features in the prior dataset in complaint work orders of different evaluation types, and the second statistical value of the number of times the business features occur in complaint work orders of the evaluation type, and determine the class conditional probability based on the first statistical value and the second statistical value.

[0113] In some embodiments, the apparatus 500 for evaluating the quality of complaint work orders may further include: The extraction module is used to extract feature values ​​for each target business feature from the complaint work order data to be evaluated; The calculation module is also used to determine the class conditional probability corresponding to the feature value for each complaint work order evaluation type, and multiply the class conditional probabilities corresponding to all target business features to obtain the joint conditional probability under that type. The calculation module is also used to multiply the joint conditional probability with the prior probability of the complaint work order type to obtain the posterior probability that the complaint work order to be evaluated belongs to different complaint work order evaluation types.

[0114] Figure 5 The various modules in the illustrated device can achieve Figure 1 The various steps involved, and the corresponding technical effects achieved, will not be elaborated upon here for the sake of brevity.

[0115] Figure 6 A schematic diagram of the hardware structure of the terminal device provided in an embodiment of this application is shown.

[0116] The terminal device may include a processor 601 and a memory 602 storing computer program instructions.

[0117] Specifically, the processor 601 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0118] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 602 may include removable or non-removable (or fixed) media, or memory 602 may be non-volatile solid-state memory. Memory 602 may be internal or external to the integrated gateway disaster recovery device.

[0119] In one instance, memory 602 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Thus, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the complaint ticket quality assessment method according to this disclosure.

[0120] The processor 601 reads and executes computer program instructions stored in the memory 602 to achieve... Figure 1 The method for evaluating the quality of complaint work orders in the illustrated embodiment.

[0121] In one example, the terminal device may further include a communication interface 603 and a bus 604. Wherein, for example... Figure 6 As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 604 and complete communication with each other.

[0122] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0123] Bus 604 includes hardware, software, or both, that couples components of an end device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 604 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0124] Furthermore, in conjunction with the complaint ticket quality assessment method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the complaint ticket quality assessment methods in the above embodiments.

[0125] This application also provides a computer program product, including a computer program, which, when executed, implements any of the complaint work order quality assessment methods described in the above embodiments.

[0126] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0127] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or text segments used to perform the required tasks. Programs or text segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Text segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0128] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0129] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0130] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for evaluating the quality of complaint work orders, characterized in that, The method includes: The system obtains the evaluation type of complaint tickets for communication services, determines the prior probability based on the statistical value of the evaluation type of complaint ticket data from historical complaint ticket data, and determines the class-conditional probability based on the statistical value of the target service feature from historical complaint ticket data; wherein, the target service feature is a service feature in which the expected value of the variance inflation factor in historical complaint ticket data is greater than a set threshold. Obtain the complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

2. The method for quality assessment of complaint work orders according to claim 1, characterized in that, Before obtaining the prior probability of the complaint ticket evaluation type for communication services, determined based on the statistical values ​​of the complaint ticket evaluation types from historical complaint ticket data, and the class-conditional probability, determined based on the statistical values ​​of the target service characteristics from historical complaint ticket data, the method further includes: Acquire multiple historical complaint ticket data and feature information of historical complaint tickets for communication services. The feature information includes multiple service features and preset correlation data of the service features. The weight of each business feature is determined based on the preset correlation data of multiple business features, and the expected value of the variance inflation factor of each business feature is determined based on historical complaint work order data, business features and weights. The business features whose expected values ​​are greater than a set threshold are identified as target business features, and a business feature template is generated based on the target business features; A prior dataset is generated based on the business feature template and historical complaint work order data, and the prior probability and class conditional probability of the complaint work order evaluation type are obtained by statistically analyzing the data in the prior dataset.

3. The method for quality assessment of complaint work orders according to claim 2, characterized in that, The weight of each business feature is determined based on preset correlation data of multiple business features, including: Based on preset correlation data of multiple business features, the number of business features that are not related to each other in terms of business is counted; the first business feature is any one of the multiple business features, and the second business feature is any business feature other than the first business feature. Calculate the ratio of the number of business features that are not related to the second business feature to the total number of business features for each first business feature, and determine the ratio as the weight of the first business feature.

4. The method for quality assessment of complaint work orders according to claim 2, characterized in that, The expected value of the variance inflation factor for each business feature is determined based on historical complaint work order data, business characteristics, and weights, including: The historical complaint work order data is split into multiple data subsets; For each data subset, regression analysis is performed with the second business feature as the independent variable and the first business feature as the dependent variable to obtain the determination coefficient of the first business feature; the first business feature is any one of multiple business features, and the second business feature is any business feature other than the first business feature. The variance inflation factor of the first business feature corresponding to each data subset is determined based on the determination coefficient of the first business feature and the weight of the business feature. Calculate the mean of multiple variance inflation factors for all data subsets of each first business feature to obtain the expected value of the variance inflation factor for the first business feature.

5. The method for quality assessment of complaint work orders according to claim 4, characterized in that, The step of performing regression analysis based on each data subset, using the second business feature as the independent variable and the first business feature as the dependent variable, to obtain the determination coefficient of the first business feature includes: Establish a linear relationship model with the second business feature as the independent variable and the first business feature as the dependent variable; The coefficients of each independent variable in the linear relationship model are determined based on the data in the current data subset, so that the difference between the actual value of the dependent variable and the predicted value obtained based on the linear relationship model is less than a set threshold. Based on the coefficients corresponding to each independent variable and the linear relationship model, the predicted value is calculated, and the coefficient of determination is calculated according to the degree of difference between the actual value of the dependent variable and the predicted value.

6. The method for quality assessment of complaint work orders according to claim 2, characterized in that, The prior probability and class-conditional probability of the complaint work order evaluation type are obtained by statistically analyzing the data in the prior dataset, including: The ratio of the number of complaint work orders of different evaluation types in the prior dataset to the total number of complaint work orders is calculated to obtain the initial prior probabilities of different complaint work order evaluation types. The prior probabilities of different complaint work order evaluation types are determined based on the initial prior probabilities of the complaint work order evaluation types and the preset weights of the complaint work order evaluation types. The first statistical value of the occurrence of the business feature in the prior dataset in complaint work orders of different evaluation types, and the second statistical value of the number of times the business feature occurs in complaint work orders of the same evaluation type are calculated respectively, and the class conditional probability is determined based on the first statistical value and the second statistical value.

7. The method for quality assessment of complaint work orders according to claim 1, characterized in that, The posterior probability of the complaint work order to be evaluated is determined based on the prior probability, class conditional probability, and the data of the complaint work order to be evaluated, including: Extract the feature values ​​of each target business feature from the complaint work order data to be evaluated; For each complaint work order evaluation type, the class conditional probability corresponding to the feature value is determined, and the class conditional probabilities corresponding to all target business features are multiplied to obtain the joint conditional probability under that type. Multiplying the joint conditional probability by the prior probability of the complaint ticket type yields the posterior probability that the complaint ticket to be evaluated belongs to different complaint ticket evaluation types.

8. A device for quality assessment of complaint work orders, characterized in that, The device includes: The acquisition module is used to acquire the evaluation type of the complaint work order for communication services, the prior probability determined based on the statistical value of the evaluation type of the complaint work order based on the historical complaint work order data, and the class conditional probability determined based on the statistical value of the target service feature based on the historical complaint work order data; wherein, the target service feature is the service feature in which the expected value of the variance inflation factor in the historical complaint work order data is greater than a set threshold. The determination module is used to acquire complaint work order data to be evaluated, determine the posterior probability of the complaint work order to be evaluated based on the prior probability, class conditional probability and the complaint work order data to be evaluated, and determine the quality classification of the complaint work order to be evaluated based on the posterior probability.

9. A terminal device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the method for quality assessment of complaint work orders as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the method for quality assessment of complaint work orders as described in any one of claims 1-7.

11. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the complaint work order quality assessment method as described in any one of claims 1-7.

Images