Complaint hotspot problem monitoring and early warning method and device and readable storage medium

Abstract

The invention provides a complaint hotspot problem monitoring and early warning method and device and a readable storage medium. The method comprises the steps of obtaining a complaint work order data set; according to the complaint work order data set, a keyword library is constructed, and the keyword library comprises conventional question keywords and personalized question keywords; according to the complaint work order data set, carrying out clustering analysis based on the city label and the hot question label to obtain a complaint quantity statistical result of different question classification; according to the complaint work order data set, the keyword library and the complaint quantity statistical result, performing prediction through a Prophet time sequence prediction model to obtain a complaint prediction result; and when the real-time complaint amount exceeds the complaint prediction result, outputting complaint early warning information. According to the invention, the method achieves the automatic processing of the complaint work order, and achieves the real-time monitoring of a hotspot problem. Meanwhile, in the process of processing complaint work order problems, personalized processing of actual problems is achieved, and prediction and early warning deviations are reduced.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a method, device and readable storage medium for monitoring and early warning of hot complaint issues. Background Art

[0002] With the rapid development of communication technology, users' requirements for communication use are getting higher and higher. Optimizing communication services has become the best choice for communication operators to improve user experience satisfaction and loyalty.

[0003] In the process of telecommunication operations, user complaints are a common phenomenon. How to respond to customer complaints promptly and quickly has become an important part of improving user satisfaction. To address this phenomenon, the current method is to mine relevant information about user experience from a large amount of complaint ticket data, and to achieve user experience evaluation, real-time monitoring, feedback, and early warning through model construction.

[0004] However, the existing complaint warning methods have many shortcomings, such as being unable to monitor hot issues in real time and the accuracy of warning results is low. Summary of the invention

[0005] The technical problem to be solved by the present application is to provide a method, device and readable storage medium for monitoring and early warning of hot complaint issues in response to the above-mentioned deficiencies in the prior art, so as to solve the problems existing in the prior art.

[0006] In a first aspect, the present application provides a method for monitoring and early warning of hot complaint issues, the method comprising:

[0007] S1. Obtain complaint work order dataset;

[0008] S2. Construct a keyword thesaurus based on the complaint ticket dataset, wherein the keyword thesaurus includes common problem keywords and personalized problem keywords;

[0009] S3. Based on the complaint work order data set, cluster analysis is performed based on city labels and hot issue labels to obtain statistical results of the number of complaints classified by different issues;

[0010] S4. According to the complaint work order dataset, the keyword vocabulary and the complaint quantity statistics, a Prophet time series prediction model is used to perform prediction to obtain a complaint prediction result, wherein the complaint prediction result includes a complaint quantity prediction threshold corresponding to different problem classifications and / or different cities in different time periods;

[0011] S5. When the real-time complaint volume exceeds the complaint prediction result, complaint warning information is output.

[0012] In some embodiments, in S2, the general question keywords and / or the individual question keywords are obtained through manual annotation based on the complaint work order dataset.

[0013] In some embodiments, in S2, the process of obtaining the personality question keywords includes:

[0014] Based on the complaint ticket dataset, the BiLSTM model is used to extract text semantic features;

[0015] Based on the semantic features of the text, the CRF probability model is used to predict labels and obtain the complaint object and complaint opinion;

[0016] The complaint objects and complaint opinions are associated and combined to obtain individual problem keywords.

[0017] In some embodiments, after combining the complaint object and the complaint opinion, the following is further included:

[0018] The similarity of the obtained personality question keywords is calculated by a similarity algorithm, and the personality question keywords whose similarity reaches a threshold are merged.

[0019] In some embodiments, S4 includes:

[0020] According to the complaint work order data set, the keyword word library and the complaint quantity statistics, the Prophet time series forecasting model data is decomposed to obtain trend items, season items and holiday items; wherein the trend item is used to characterize the long-term dynamic evolution trend of the time series, the season item is used to characterize the periodic fluctuation law inherent in the time series, and the holiday item is used to characterize the sudden fluctuation caused by holidays in the time series;

[0021] Based on the trend items, seasonal items and holiday items, the Prophet time series forecasting model is used to perform forecasting to obtain complaint forecasting results.

[0022] In some embodiments, based on the trend item, season item and holiday item, the Prophet time series prediction model is used to perform prediction according to the following formula to obtain a complaint prediction result:

[0023] g(t)=k*t+m

[0024]

[0025] y(t)=g(t)+s(t)+h(t)

[0026] Among them, g(t) is the trend term, k is the slope, m is the intercept, t is the time variable in the model, s(t) is the seasonal term, N is the order of the Fourier series, P is the period length of the seasonal pattern, and an and b n is the Fourier coefficient, h(t) is the holiday term, and y(t) is the complaint prediction result.

[0027] In some embodiments, the complaint work order data set includes at least one of the work order serial number, business number, complaint location code, complaint city, complaint acceptance time, settlement time, customer complaint description, acceptance channel, complaint problem location, responsible profession, customer level, business category, package name, handling opinions, and account period.

[0028] In a second aspect, the present application provides a device for monitoring and early warning of hot complaint issues, the device comprising:

[0029] A data acquisition module, which is configured to acquire a complaint ticket data set;

[0030] A word library construction module, which is configured to construct a keyword word library based on the complaint work order data set, wherein the keyword word library includes common problem keywords and individual problem keywords;

[0031] A cluster analysis module is configured to perform cluster analysis based on the complaint work order data set, city labels and hot issue labels, to obtain statistical results of the number of complaints classified by different issues;

[0032] A complaint prediction module is configured to perform prediction based on the complaint work order data set, the keyword vocabulary and the complaint quantity statistics through the Prophet time series prediction model to obtain complaint prediction results, wherein the complaint prediction results include complaint quantity prediction thresholds corresponding to different problem classifications and / or different cities in different time periods;

[0033] The complaint warning module is configured to output complaint warning information when the real-time complaint volume exceeds the complaint prediction result.

[0034] In a third aspect, the present application provides a device for monitoring and early warning of hot complaint issues, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the method for monitoring and early warning of hot complaint issues described in the first aspect above.

[0035] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for monitoring and early warning of hot complaint issues described in the first aspect above is implemented.

[0036] The present application provides a method, device and readable storage medium for monitoring and warning of hot complaint issues. The method includes: obtaining a complaint work order data set; constructing a keyword thesaurus based on the complaint work order data set, wherein the keyword thesaurus includes common problem keywords and individual problem keywords; performing cluster analysis based on the complaint work order data set and the city labels and the hot problem labels to obtain the statistical results of the number of complaints for different problem classifications; performing predictions through the Prophet time series prediction model based on the complaint work order data set, the keyword thesaurus and the statistical results of the number of complaints to obtain complaint prediction results, wherein the complaint prediction results include the prediction thresholds of the number of complaints corresponding to different problem classifications and / or different cities in different time periods; when the real-time complaint volume exceeds the complaint prediction results, outputting complaint warning information. The present application realizes the automated processing of complaint work orders, and achieves the effect of real-time monitoring of hot issues. At the same time, in the process of processing complaint work order issues, they are divided into common issues and individual issues, so as to achieve personalized processing of actual problems and reduce prediction and warning deviations. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0038] Figure 1 A flowchart of a method for monitoring and early warning of hot complaint issues provided in an embodiment of the present application;

[0039] Figure 2 A flowchart of another method for monitoring and early warning of hot complaint issues provided in an embodiment of the present application;

[0040] Figure 3 A schematic diagram of a process for obtaining keywords for personality questions provided in an embodiment of the present application;

[0041] Figure 4 A schematic diagram of the structure of a hot complaint monitoring and early warning device provided in an embodiment of the present application;

[0042] Figure 5 A schematic diagram of the structure of another complaint hotspot monitoring and early warning device provided in an embodiment of the present application.

[0043] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the technical solution of the present application, the implementation mode of the present application will be further described in detail below with reference to the accompanying drawings.

[0045] It should be understood that the specific embodiments and drawings described herein are only used to explain the present application, rather than to limit the present application.

[0046] It can be understood that, in the absence of conflict, the various embodiments in the present application and the various features in the embodiments can be combined with each other.

[0047] It can be understood that, for the convenience of description, the drawings of the present application only show the parts related to the present application, while the parts unrelated to the present application are not shown in the drawings.

[0048] It can be understood that each unit and module involved in the embodiments of the present application may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0049] It can be understood that the terms "first", "second", etc. in the embodiments of the present application are used to distinguish different objects, or to distinguish different processing of the same object, rather than to describe a specific order of objects.

[0050] It is understandable that, in the absence of conflict, the functions and steps marked in the flowcharts and block diagrams of the present application may occur in an order different from that marked in the drawings.

[0051] It is understood that the flowcharts and block diagrams of the present application illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present application. Among them, each box in the flowchart or block diagram may represent a unit, module, program segment, code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or by a combination of hardware and computer instructions.

[0052] It can be understood that the units and modules involved in the embodiments of the present application can be implemented by software or hardware, for example, the units and modules can be located in a processor.

[0053] In the prior art, existing methods for hot issue monitoring and early warning research can be mainly divided into the following two categories:

[0054] (1) Construction of complaint ticket classification model

[0055] There are two main methods for building complaint ticket classification models. One is to build a keyword thesaurus for clustering statistical analysis, and the other is to use feature vectors for machine learning. The former is based on natural language processing algorithms to extract entities from historical complaint ticket text data, build a keyword thesaurus for hot issues, and cluster different keywords to obtain multiple hot event types. The latter builds a data mining model, uses the complaint ticket text information as the model input, and uses the ticket text data with category labels as the model output to build a supervised machine learning model.

[0056] (2) Complaint prediction and early warning:

[0057] The existing technology uses the correlation analysis method, which first determines the target warning information set and the complaint content set, then associates the complaint content with the target warning information, and finally counts the target complaint work order number corresponding to each target warning information, that is, the number of complaints. When the number of complaints meets the threshold condition, an early warning is triggered.

[0058] However, the existing methods for monitoring and early warning of hot issues in complaint tickets have the following shortcomings:

[0059] 1) Currently, most methods only study and build models for the classification of complaint tickets or complaint prediction and early warning, but there is no comprehensive hot issue classification and early warning solution, and the full automation of complaint ticket processing has not been achieved. Especially with the rapid development of communication technology, the number of complaint tickets generated daily increases, and semi-automatic processing methods will lead to many problems such as low processing efficiency, long processing time, inaccurate processing results, and inability to monitor hot issues in real time.

[0060] 2) Most models for processing complaint tickets classify all tickets into one category, and are unable to establish a more refined and personalized data mining model for complaint tickets, resulting in deviations in real-time monitoring: Currently, there are two main types of problems reflected in user complaints. One is routine problems, such as clarification of call charges, mobile network coverage, etc., and the other is customer feedback caused by new activities and new policies currently being carried out. The two types of problems have slightly different requirements for data support and processing procedures, but existing methods only classify them into one category for analysis, which will cause errors.

[0061] In view of the shortcomings of the above existing technologies, this application innovatively develops a model for monitoring and warning of hot complaint issues. The model integrates the classification model and warning model of complaint tickets into the same processing flow, thereby realizing the automated processing of complaint tickets and achieving the effect of real-time monitoring of hot issues. At the same time, in the process of handling complaint tickets, they are divided into routine issues and individual issues, achieving personalized processing of actual issues and reducing prediction and warning deviations.

[0062] This application establishes a monitoring and early warning model for hot complaint issues based on the complaint ticket text data. The overall logical thinking of model construction can be divided into five steps:

[0063] First, obtain the complaint ticket dataset from the database of the new customer service management system;

[0064] Secondly, we build a keyword database for complaint issues. The construction process of this database is to divide complaint issues into individual issues and general issues. For general issues, we build a keyword database for hot issues based on historical complaint ticket data. For individual issues, we can find complaint keywords through two methods: manual active configuration and automatic recognition.

[0065] Then, based on the keyword database results from the previous step, cluster statistical analysis was performed from the perspectives of cities and hot issue categories to obtain quantitative statistics under different classification conditions;

[0066] Next, build a Prophet time series forecasting model to predict the number of complaint tickets in the next hour / day;

[0067] Finally, compare the time series prediction number with the real-time generation number. If the actual number of complaint tickets is greater than the time series prediction number, an early warning will be issued for hot issues.

[0068] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0069] The present application provides a method for monitoring and early warning of hot complaint issues. The working process of the method can be implemented by electronic devices, such as computers, handheld smart terminals, etc. For the convenience of explanation, the embodiments of the present application are described with the method execution subject being a computer.

[0070] Figure 1 A schematic diagram of a method for monitoring and early warning of hot complaint issues provided in an embodiment of the present application. Figure 2 Another schematic diagram of the method for monitoring and early warning of hot complaint issues provided in the embodiment of the present application is as follows: Figure 1 as well as Figure 2 As shown, the present application provides a method for monitoring and early warning of hot complaint issues, the method includes S1-S5, which are as follows:

[0071] S1. Obtain complaint work order dataset;

[0072] In some embodiments, the complaint work order data set includes at least one of the work order serial number, business number, complaint location code, complaint city, complaint acceptance time, settlement time, customer complaint description, acceptance channel, complaint problem location, responsible profession, customer level, business category, package name, handling opinions, and account period.

[0073] Specifically, based on Kafka technology, the complaint ticket dataset is obtained from the database of the new customer service system. The fields of the original complaint ticket dataset include ticket serial number, business number, complaint location code, complaint city, complaint acceptance time, settlement time, customer complaint description, acceptance channel, complaint problem location, responsible profession, customer level, business category, package name, handling opinion, account period, etc. This original dataset serves as the basis for subsequent processing.

[0074] S2. Construct a keyword thesaurus based on the complaint ticket dataset, wherein the keyword thesaurus includes common problem keywords and personalized problem keywords;

[0075] In this step, user complaints are divided into two main parts: one part is routine problems, such as telephone charges, mobile network coverage, etc., and the other part is unconventional individual problems, such as customer feedback caused by new activities and new policies currently being carried out.

[0076] In some embodiments, in S2, the general question keywords and / or the individual question keywords are obtained through manual annotation based on the complaint work order dataset.

[0077] Specifically, for common problems, the historical complaint ticket data content is manually annotated to find keyword entries for different types of hot issues, and finally form a keyword knowledge base for common problems.

[0078] In addition, for individual problems, it depends on new complaint text data. When the keyword information of new communication packages is known, manual summary configuration can be performed to obtain keyword entries for individual problems.

[0079] In some embodiments, in S2, the process of obtaining the personality question keywords includes:

[0080] Based on the complaint ticket dataset, the BiLSTM model is used to extract text semantic features;

[0081] Based on the semantic features of the text, the CRF probability model is used to predict labels and obtain the complaint object and complaint opinion;

[0082] The complaint objects and complaint opinions are associated and combined to obtain individual problem keywords.

[0083] In some embodiments, after combining the complaint object and the complaint opinion, the following is further included:

[0084] The similarity of the obtained personality question keywords is calculated by a similarity algorithm, and the personality question keywords whose similarity reaches a threshold are merged.

[0085] In the process of acquiring individual problem keywords, for sudden complaint events with unknown keyword information, natural language processing is performed on the complaint ticket text dataset based on the new complaint ticket data. The text data of the complaint ticket is segmented by combining the commonly used Chinese stop word library and the constructed keyword vocabulary. At the same time, non-keywords are cleaned to obtain a complaint ticket text dataset with category labels, so as to perform automatic recognition to identify keyword entries that have not been considered.

[0086] Specifically, the automatic recognition algorithm includes: Based on the complaint text data received in real time, the BiLSTM+CRF algorithm is used. It is an efficient method that can accurately segment the text and identify the key entities and the relationship between these entities. This technology is suitable for extracting structured information from unstructured customer feedback, such as identifying the object of the complaint (such as products, personnel, etc.) and the specific point of view of the complaint (such as inaccurate billing, poor service attitude, etc.). Then, cosine similarity calculation is used to merge similar problems and extract new personalized problems that have not been considered.

[0087] Among them, BiLSTM (Bidirectional Long Short-Term Memory) is an extension of the Long Short-Term Memory (LSTM). LSTM is a special recurrent neural network (RNN) that can learn long-term dependency information. BiLSTM combines two LSTMs, one for processing forward sequences and the other for processing reverse sequences.

[0088] CRF (Conditional Random Field) is a probabilistic graphical model that is often used for labeling problems of sequence data, such as part-of-speech tagging and named entity recognition in natural language processing.

[0089] Figure 3 A schematic diagram of the process of obtaining the personality question keywords provided in the embodiment of the present application is shown in FIG. Figure 3 As shown, the process mainly includes the following steps:

[0090] (1) Data preprocessing

[0091] First, the received complaint text is cleaned, such as removing irrelevant characters and punctuation marks, converting uppercase and lowercase letters, and performing word segmentation.

[0092] (2) Extracting text semantic features

[0093] At this stage, our goal is to extract useful features from the text so that the model can better understand the text content. Here, the BiLSTM model is mainly used to extract semantic features in the text.

[0094] The BiLSTM model is divided into forward LSTM and backward LSTM. The forward LSTM reads the text from front to back and captures the context information from left to right; the backward LSTM reads the text from back to front and captures the context information from right to left. The outputs of the forward and backward LSTM are combined to form the final representation of each word.

[0095] By combining contextual information in two directions, the BiLSTM model is able to better understand the meaning of a word in a sentence and its relationship with other words.

[0096] (3) Sequence labeling to identify the object of complaint and the complaint opinion

[0097] CRF is a probabilistic model for sequence labeling, which can take into account the label distribution of the entire sequence and ensure the consistency and rationality of the label sequence. Adding a CRF layer after BiLSTM can predict the label for each word.

[0098] A tag set typically contains:

[0099] O: Non-critical information

[0100] B-Aspect: Object of complaint

[0101] B-Opinion: Complaint Opinion

[0102] The model predicts a label for each word, thus obtaining a label sequence for the entire sentence.

[0103] (4) Recognition result analysis and association

[0104] The complaint objects and complaint opinions identified by the model are combined to form the final result, such as inaccurate product billing and poor service attitude of personnel.

[0105] Optionally, in order to reduce the interference of synonyms on result aggregation, for example, if inaccurate billing and billing doubt express the same problem, the cosine similarity is used to calculate the similarity of the recognition results, and the ones with high similarity are merged into the same problem.

[0106] The calculation formula for cosine similarity is:

[0107]

[0108] Wherein, A·B represents the dot product of result A and result B after conversion to vectors; ||A|| and ||B|| represent the modulus of result A and result B after conversion to vectors, respectively.

[0109] For the final recognition results, if there are complaint issues that do not appear in the knowledge base, they will be identified as new personal issues.

[0110] Through the processing of step S2, the common problem keywords and individual problem keywords finally obtained jointly construct the keyword vocabulary of the complaint problem.

[0111] S3. Based on the complaint work order data set, cluster analysis is performed based on city labels and hot issue labels to obtain statistical results of the number of complaints classified by different issues;

[0112] Specifically, for the structured complaint ticket text data, cluster analysis is performed on the text data based on city labels and hot issue labels, and the number of complaints of each type is counted to provide a prediction data basis for the subsequent time series prediction algorithm.

[0113] In this step, the clustering of prefecture-level cities is based on the names of prefecture-level cities, and the clustering results of hot issue categories (showing the top 5 categories) are shown in the following table:

[0114]

[0115]

[0116] S4. According to the complaint work order dataset, the keyword vocabulary and the complaint quantity statistics, a Prophet time series prediction model is used to perform prediction to obtain a complaint prediction result, wherein the complaint prediction result includes a complaint quantity prediction threshold corresponding to different problem classifications and / or different cities in different time periods;

[0117] In some embodiments, S4 includes:

[0118] According to the complaint work order data set, the keyword word library and the complaint quantity statistics, the Prophet time series forecasting model data is decomposed to obtain trend items, season items and holiday items; wherein the trend item is used to characterize the long-term dynamic evolution trend of the time series, the season item is used to characterize the periodic fluctuation law inherent in the time series, and the holiday item is used to characterize the sudden fluctuation caused by holidays in the time series;

[0119] Based on the trend items, seasonal items and holiday items, the Prophet time series forecasting model is used to perform forecasting to obtain complaint forecasting results.

[0120] Specifically, the time series algorithms for predicting future trends based on historical data include prophet and ARIMA algorithms, but experiments show that the root mean square error obtained by the prophet algorithm is smaller than that of the ARIMA algorithm, so the prophet time series algorithm is used for time series prediction of cities and problem categories. The algorithm can analyze time series of multiple time periods (comparison of a certain day in the same week, comparison of the same day in different weeks, comparison of weekends and weekdays) to form function rules.

[0121] Prophet is a decomposition model that decomposes time series data into three components: trend, seasonality, and holiday effects. Prophet fits these components through linear regression and predicts future trends and seasonal patterns. The following are the explanations of each part:

[0122] Trend term: reveals and captures the long-term dynamic evolution trend of the time series, and uses a piecewise linear model with auto-regression characteristics and seasonal adjustment for fitting. This trend can be expressed as linear development or flexibly adapt to nonlinear trends, depending entirely on the actual data characteristics.

[0123] Seasonal term: In-depth analysis of the inherent cyclical fluctuations in time series, such as repetitive patterns at different levels of cycles such as days, weeks, and even years. Prophet cleverly uses the Fourier series method to accurately simulate and characterize these seasonal effects.

[0124] Holiday items: sensitively identify and respond to sudden fluctuations in time series caused by holidays, promotions or other special events. Prophet supports user-defined holiday events and can set a specific impact model for each holiday to quantify the increase or decrease in the number of work orders.

[0125] In some embodiments, based on the trend item, season item and holiday item, the Prophet time series prediction model is used to perform prediction according to the following formula to obtain a complaint prediction result:

[0126] g(t)=k*t+m

[0127]

[0128] y(t)=g(t)+s(t)+h(t)

[0129] Among them, g(t) is the trend term, k is the slope, m is the intercept, k and m are obtained by the model through fitting data training, t is the time variable in the model, s(t) is the seasonal term, N is the order of the Fourier series, which determines the complexity of the Fourier series expansion. The default setting of Prophet is 10, and P is the cycle length of the seasonal pattern, such as daily seasonality P = 1, weekly seasonality P = 7, an and b n is the Fourier coefficient, which is obtained by the model through fitting data training, h(t) is the holiday term, which is obtained by specifying a specific date and the model is trained by fitting the data on the specified date, and y(t) is the complaint prediction result.

[0130] Based on the above calculation formula, the prophet algorithm finally generates the predicted value of the future trend of the time series through the organic integration and superposition calculation of the trend item, seasonal item and holiday item. In addition, this algorithm can increase the prediction accuracy of the time series by adaptively optimizing as the real-time time series data increases.

[0131] S5. When the real-time complaint volume exceeds the complaint prediction result, complaint warning information is output.

[0132] Specifically, the predicted value obtained by prophet time series prediction is used as the warning threshold of the number of complaint tickets. When the actual number of complaint tickets exceeds the threshold, it can be considered that there is a risk of hidden dangers in the number of complaints, and it needs to be dealt with in a timely manner to prevent larger operational complaint problems.

[0133] For example, early warning according to the city clustering perspective is shown in the following table:

[0134] City Prediction Interval Real-time complaint volume Nantong City <＝13 18

[0135] Among them, the threshold 13 in the prediction interval is obtained based on the time series prediction in step 4, and the early warning formula is:

[0136]

[0137] Here, x refers to the real-time complaint volume. When the real-time complaint ticket volume exceeds the threshold of 13, the system provides early warning information to the terminal personnel.

[0138] For another example, early warnings based on hot issues are shown in the following table:

[0139]

[0140] Among them, threshold 1 is also obtained based on time series prediction, and the warning method formula is similar to the warning formula based on the city perspective, that is:

[0141]

[0142] When the number of real-time complaint work orders exceeds the threshold of 1, the system provides early warning information to terminal personnel.

[0143] This application provides a method for monitoring and early warning of hot complaint issues, which actively identifies hot spots and individual issues in complaint work orders. By analyzing the text data of historical complaint work orders, using natural language processing (NLP) technology to accurately segment the content of new work orders, the BiLSTM+CRF algorithm is used to actively and efficiently identify common and individual issues, and finally improve the construction of the keyword vocabulary. In addition, this application integrates the natural language processing algorithm and the prophet time series prediction algorithm, which can efficiently identify common and individual issues, and warn of cities and hot issues that exceed the algorithm prediction threshold. It greatly improves the efficiency of traditional manual work order processing and improves the process processing speed of automatic early warning of complaint work orders.

[0144] This application establishes a monitoring and early warning model based on hot issues, and uses natural language processing algorithms (including BiLSTM+CRF+cosine similarity and other technologies) and time series prediction algorithms to achieve efficient identification and effective early warning of complaint work order issues, and ultimately achieve automated processing of complaint work order issues.

[0145] It should be understood that, although the various steps in the flowcharts in the above-described embodiments are sequentially displayed according to the indications of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least a portion of the steps in the figure may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily to be carried out sequentially, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0146] Figure 4 A schematic diagram of a hot complaint monitoring and early warning device provided in an embodiment of the present application, such as Figure 4 As shown, the present application provides a complaint hotspot monitoring and early warning device. In some embodiments, the device includes:

[0147] A data acquisition module 11, which is configured to acquire a complaint work order data set;

[0148] A word library construction module 12, which is configured to construct a keyword word library according to the complaint work order data set, wherein the keyword word library includes common problem keywords and individual problem keywords;

[0149] A cluster analysis module 13 is configured to perform cluster analysis based on the complaint work order data set, city labels and hot issue labels, and obtain statistical results of the number of complaints classified by different issues;

[0150] The complaint prediction module 14 is configured to perform prediction based on the complaint work order data set, the keyword vocabulary and the complaint quantity statistics through the Prophet time series prediction model to obtain a complaint prediction result, wherein the complaint prediction result includes a complaint quantity prediction threshold corresponding to different problem classifications and / or different cities in different time periods;

[0151] The complaint warning module 15 is configured to output complaint warning information when the real-time complaint volume exceeds the complaint prediction result.

[0152] Regarding the limitation of the hot complaint problem monitoring and early warning device, reference may be made to the limitation of the hot complaint problem monitoring and early warning method in the above-mentioned embodiments of the present application, which will not be elaborated in this embodiment.

[0153] Figure 5 Another schematic diagram of the complaint hotspot monitoring and early warning device provided in the embodiment of the present application is as follows: Figure 5 As shown, in some embodiments, the present application provides a complaint hotspot monitoring and early warning device, including a memory 22 and a processor 21, the memory stores a computer program, and the processor is configured to run the computer program to execute the complaint hotspot monitoring and early warning methods in the above-mentioned embodiments of the present application.

[0154] The memory is connected to the processor, the memory may be a flash memory or a read-only memory or other memory, and the processor may be a central processing unit or a single-chip microcomputer.

[0155] In some embodiments, the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for monitoring and early warning of hot complaint issues in the above-mentioned embodiments of the present application is implemented.

[0156] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0157] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present application, but the present application is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and substance of the present application, and these modifications and improvements are also considered to be within the scope of protection of the present application.

Claims

1. A method for monitoring and early warning of hot complaint issues, characterized in that: The method comprises: S1. Obtain complaint work order dataset; S2. Construct a keyword thesaurus based on the complaint ticket dataset, wherein the keyword thesaurus includes common problem keywords and personalized problem keywords; S3. Based on the complaint work order data set, cluster analysis is performed based on city labels and hot issue labels to obtain statistical results of the number of complaints classified by different issues; S4. According to the complaint work order dataset, the keyword vocabulary and the complaint quantity statistics, a Prophet time series prediction model is used to perform prediction to obtain a complaint prediction result, wherein the complaint prediction result includes a complaint quantity prediction threshold corresponding to different problem classifications and / or different cities in different time periods; S5. When the real-time complaint volume exceeds the complaint prediction result, complaint warning information is output.

2. The method for monitoring and early warning of hot complaint issues according to claim 1 is characterized in that: In S2, the conventional question keywords and / or the individual question keywords are obtained through manual annotation based on the complaint work order dataset.

3. The method for monitoring and early warning of hot complaint issues according to claim 1 is characterized in that: In S2, the process of obtaining the personality question keywords includes: Based on the complaint ticket dataset, the BiLSTM model is used to extract text semantic features; Based on the semantic features of the text, the CRF probability model is used to predict labels and obtain the complaint object and complaint opinion; The complaint objects and complaint opinions are associated and combined to obtain individual problem keywords.

4. The method for monitoring and early warning of hot complaint issues according to claim 3 is characterized in that: After combining the object of complaint and the complaint opinion, it also includes: The similarity of the obtained personality question keywords is calculated by a similarity algorithm, and the personality question keywords whose similarity reaches a threshold are merged.

5. The method for monitoring and early warning of hot complaint issues according to claim 1 is characterized in that: S4, including: According to the complaint work order data set, the keyword word library and the complaint quantity statistics, the Prophet time series forecasting model data is decomposed to obtain trend items, season items and holiday items; wherein the trend item is used to characterize the long-term dynamic evolution trend of the time series, the season item is used to characterize the periodic fluctuation law inherent in the time series, and the holiday item is used to characterize the sudden fluctuation caused by holidays in the time series; Based on the trend items, seasonal items and holiday items, the Prophet time series forecasting model is used to perform forecasting to obtain complaint forecasting results.

6. The method for monitoring and early warning of hot complaint issues according to claim 5 is characterized in that: Based on the trend item, seasonal item and holiday item, the Prophet time series forecasting model is used to perform forecasting according to the following formula to obtain the complaint forecast result: g(t)=k*t+m y(t)=g(t)+s(t)+h(t) Among them, g(t) is the trend term, k is the slope, m is the intercept, t is the time variable in the model, s(t) is the seasonal term, N is the order of the Fourier series, P is the period length of the seasonal pattern, and a n and b n is the Fourier coefficient, h(t) is the holiday term, and y(t) is the complaint prediction result.

7. The method for monitoring and early warning of hot complaint issues according to any one of claims 1 to 6, characterized in that: The complaint work order data set includes at least one of the work order serial number, business number, complaint location code, complaint city, complaint acceptance time, settlement time, customer complaint description, acceptance channel, complaint problem location, responsible profession, customer level, business category, package name, handling opinion, and account period.

8. A device for monitoring and early warning of hot complaint issues, characterized in that: The device comprises: A data acquisition module, which is configured to acquire a complaint ticket data set; A word library construction module, which is configured to construct a keyword word library based on the complaint work order data set, wherein the keyword word library includes common problem keywords and individual problem keywords; A cluster analysis module is configured to perform cluster analysis based on the complaint work order data set, city labels and hot issue labels, to obtain statistical results of the number of complaints classified by different issues; A complaint prediction module is configured to perform prediction based on the complaint work order data set, the keyword vocabulary and the complaint quantity statistics through the Prophet time series prediction model to obtain complaint prediction results, wherein the complaint prediction results include complaint quantity prediction thresholds corresponding to different problem classifications and / or different cities in different time periods; The complaint warning module is configured to output complaint warning information when the real-time complaint volume exceeds the complaint prediction result.

9. A device for monitoring and early warning of hot issues of complaints, characterized in that: It includes a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to implement the hot complaint problem monitoring and early warning method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for monitoring and early warning of hot complaint issues as described in any one of claims 1 to 7 is implemented.

Images