Complaint prediction method and device, nonvolatile storage medium and electronic equipment

Abstract

The invention discloses a complaint prediction method and device, a nonvolatile storage medium and electronic equipment. The method comprises the steps that complaint related data of a target area is acquired, the complaint related data comprises communication data, customer complaint data and external data, the communication data comprises operation performance index data and real-time alarm information of a base station, and the external data comprises meteorological data of the target area; constructing a time sequence matrix for the complaint related data according to a time sequence; a customer behavior memory model is adopted to analyze the time sequence matrix to obtain a prediction result, the prediction result comprises the complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model used for analyzing and predicting the time sequence data. The technical problem of low customer satisfaction caused by lack of an effective complaint prediction mechanism is solved.

Description

Technical Field

[0001] The present application relates to the field of communications, and more specifically, to a complaint prediction method, device, non-volatile storage medium, and electronic device. Background Art

[0002] With the widespread application of 45G networks, users' expectations for communication service quality are constantly increasing. Once users are dissatisfied with the network quality, they are likely to complain. User complaints reflect users' real feedback on network quality and are crucial to operators. This not only affects user experience, but also damages the brand image of operators. Therefore, operators attach great importance to complaints to improve service quality and user satisfaction, but existing technologies lack an effective complaint prediction mechanism, resulting in low customer satisfaction.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a complaint prediction method, device, non-volatile storage medium and electronic device to at least solve the technical problem of low customer satisfaction caused by the lack of an effective complaint prediction mechanism.

[0005] According to one aspect of an embodiment of the present application, a complaint prediction method is provided, including: obtaining complaint-related data of a target area, the complaint-related data including communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of a base station, and the external data includes meteorological data of the target area; constructing a time series matrix for the complaint-related data in chronological order; and analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes a complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model for analyzing and predicting time series data.

[0006] Optionally, the customer behavior memory model includes an acceptance gate, an elimination gate, a storage core and an output regulator, wherein the acceptance gate is used to determine the data stored in the storage core and store the data; the elimination gate is used to determine the data to be deleted in the storage core and delete the data to be deleted; the storage core is used to generate candidate memory states based on the data stored in the storage core, wherein the candidate memory states are used to reflect the dependency relationship between the complaint behavior and the communication data and external data; the output regulator is used to determine the target memory state from the candidate memory states, and output the prediction result based on the target memory state.

[0007] Optionally, constructing a time series matrix for complaint-related data in chronological order includes: preprocessing the complaint-related data; performing weighted normalization on the preprocessed complaint-related data; and constructing a time series matrix for the weighted normalized complaint-related data in chronological order, wherein each row in the time series matrix corresponds to complaint-related data at a time point, and each column in the time series matrix corresponds to a type of complaint-related data.

[0008] Optionally, preprocessing the complaint-related data includes: determining a first data cleaning rule corresponding to the communication data in the complaint-related data, a second data cleaning rule corresponding to the customer complaint data, and a third data cleaning rule corresponding to the external data; cleaning the communication data according to the first data cleaning rule, cleaning the customer complaint data according to the second data cleaning rule, and cleaning the external data according to the third data cleaning rule.

[0009] Optionally, the first data cleaning rule includes deleting erroneous values, missing values ​​and extreme values ​​in the operating performance indicator data of the base station, and deleting the alarm information of the base station that is determined to not affect the business; the second data cleaning rule includes deleting non-target operator complaint data and non-network problem complaint data in the customer complaint data; the third data cleaning rule includes deleting missing values ​​in meteorological data.

[0010] Optionally, weighted normalization processing is performed on the pre-processed complaint-related data, including: determining the first-level weight and second-level weight corresponding to each data in the pre-processed complaint-related data, wherein the first-level weight is used to reflect the importance of the data, and the second-level weight is used to reflect the scope of influence of the data; weighted normalization processing is performed on the pre-processed complaint-related data based on the first-level weight and the second-level weight.

[0011] Optionally, after analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, the method further includes: generating a prediction display chart based on the prediction result, and displaying the prediction display chart.

[0012] According to another aspect of an embodiment of the present application, a complaint prediction device is also provided, including: a first processing module, used to obtain complaint-related data of a target area, the complaint-related data including communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of the base station, and the external data includes meteorological data of the target area; a second processing module, used to construct a time series matrix for the complaint-related data in chronological order; a third processing module, used to analyze the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model used to analyze and predict time series data.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is further provided, in which a program is stored, wherein when the program is running, a device where the non-volatile storage medium is located is controlled to execute a complaint prediction method.

[0014] According to another aspect of an embodiment of the present application, there is also provided an electronic device, including: a memory and a processor, the processor being configured to run a program stored in the memory, wherein the complaint prediction method is executed when the program is run.

[0015] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a computer program, which implements a complaint prediction method when executed by a processor.

[0016] In an embodiment of the present application, complaint-related data of a target area is obtained, and the complaint-related data includes communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of base stations, and the external data includes meteorological data of the target area; a time series matrix is ​​constructed for the complaint-related data in chronological order; the time series matrix is ​​analyzed using a customer behavior memory model to obtain prediction results, wherein the prediction results include the complaint risk level of the target area; the customer behavior memory prediction model is a prediction model method for analyzing and predicting time series data, and the purpose of accurately predicting user complaints is achieved by integrating multi-source data and using the CBMM model for prediction, thereby achieving the technical effect of strengthening service quality and improving user satisfaction, thereby solving the technical problem of low customer satisfaction caused by the lack of an effective complaint prediction mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 It is a structural schematic diagram of a computer terminal (mobile device) provided according to an embodiment of the present application;

[0019] Figure 2 It is a flowchart of a complaint prediction method provided according to an embodiment of the present application;

[0020] Figure 3 is a schematic diagram of complaint data provided according to an embodiment of the present application;

[0021] Figure 4 It is a schematic diagram of a training process of a CBMM (customer behavior memory) model provided according to an embodiment of the present application;

[0022] Figure 5 This is an example diagram of complaint prediction results provided according to an embodiment of the present application;

[0023] Figure 6 It is a structural diagram of a complaint prediction device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0027] Mobile Network: Mobile network refers to a network system that allows users to access the Internet for data transmission and communication through mobile devices (such as mobile phones, tablets, smart watches, etc.) while on the move. It mainly includes wireless communication technology and network infrastructure, such as base stations, core networks, transmission links, etc., as well as standards and protocols that support these devices (such as 2G, 3G, 4G, 5G, etc.). Mobile networks enable users to access Internet services anytime and anywhere, such as browsing the web, sending emails, making video calls, downloading applications, etc.

[0028] CBMM (Customer Behavior Memory Model) is an advanced prediction model based on time series analysis, which is used to capture and analyze user behavior patterns. The model achieves prediction through four key steps: the admission gate determines which information should be stored in the storage core; the elimination gate determines which information should be deleted from the storage core; the storage core stores the filtered information and generates candidate memory states; the output regulator determines which information should be output. CBMM captures the long-term dependencies of user behavior through storage cores and gating mechanisms, thereby more accurately predicting future complaint trends. In addition, CBMM is able to process information from different sources such as base station performance data, user complaint data, and weather data to ensure data diversity and accuracy, analyze network status in a real-time environment, and display prediction results through visualization tools to help operators discover and solve potential problems in advance, improve service quality and user satisfaction.

[0029] Mobile network user complaints: Mobile network users, when using mobile communication services, provide feedback, request solutions to problems or express dissatisfaction to mobile network operators through formal or informal channels due to dissatisfaction or questions about service quality, tariffs, package content, network coverage, customer service, etc.

[0030] In related technologies, the main problems that operators face in handling complaints are:

[0031] 1. Lack of effective advance warning mechanism: When facing network problems, operators can only respond passively after the problem is exposed, which prolongs the processing time, affects the user experience, and may cause a large number of complaints. The lack of a pre-warning system makes it difficult for operators to discover and solve problems in advance, and thus cannot effectively prevent complaints from occurring.

[0032] 2. Inefficient and costly complaint handling: When users encounter network problems, they will complain to operators. However, the traditional complaint handling process is often cumbersome and time-consuming, resulting in low handling efficiency. At the same time, in order to resolve user complaints, operators need to invest a lot of manpower, material resources and time, which increases operating costs.

[0033] 3. High rate and large number of complaints: Due to frequent network problems and declining user experience, operators often face a large number of user complaints. These complaints not only increase the workload of operators, but also damage the brand image and affect user loyalty.

[0034] 4. High operating costs of base stations: As an important part of the communication network, the operating costs of base stations have always been the focus of operators. However, due to reasons such as aging equipment and unreasonable resource allocation, base stations frequently fail and downtime is prolonged, resulting in high operating costs.

[0035] 5. User satisfaction and experience need to be improved: In the fiercely competitive communications market, user satisfaction and experience are the core competitiveness of operators. However, due to frequent network problems and untimely complaint handling, user satisfaction and experience are often seriously affected. This not only affects user loyalty, but also limits the market expansion capabilities of operators.

[0036] Existing technologies have alleviated the data imbalance problem in AI modeling by balancing sample sets, but they still have shortcomings in handling data diversity and complexity.

[0037] In order to solve the above problems, relevant solutions are provided in the embodiments of the present application, which are described in detail below.

[0038] According to an embodiment of the present application, a method embodiment of a complaint prediction method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal (or mobile device) for implementing a complaint prediction method. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0040] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the complaint prediction method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, to implement the above-mentioned complaint prediction method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0044] Under the above operating environment, the present application embodiment provides a complaint prediction method, such as Figure 2 As shown, the method comprises the following steps:

[0045] Step S202, obtaining complaint-related data of the target area, wherein the complaint-related data includes communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of the base station, and the external data includes meteorological data of the target area.

[0046] Optionally, data collection includes using the network management system provided by the equipment manufacturer and the operator's big data platform to effectively collect key performance indicators (KPIs) and real-time alarm information of base station operation, so as to fully understand the performance status of the base station. In addition, through the operator's customer service system, detailed customer complaint data can also be obtained. Figure 3 A type of complaint data is shown, such as Figure 3 As shown in the figure, these data cover key information such as the time, specific type and location of the complaint, which helps to respond quickly and optimize the service experience. At the same time, with the help of the API interface opened by the National Meteorological Administration system or third-party service providers, future weather forecasts or historical meteorological data for various cities and counties can be obtained, providing important reference for network optimization and fault prevention.

[0047] Optionally, data collection is the basis of the entire system, and its quality directly affects the effect of subsequent processing. In this embodiment, data collection mainly includes two aspects: real-time capture of network operation indicators and integration of external factor data. These steps ensure the comprehensiveness, accuracy and timeliness of the collected data.

[0048] 1. Real-time capture of network operation indicators, careful design of system interfaces and seamless integration with network management systems, and use of API interfaces or direct database connection technology to instantly obtain core operation data such as signal strength, network delay, and packet loss rate. In the technical architecture, Kafka plays the role of a message queue to ensure the instant transmission and efficient processing of data; efficient tools such as Flume or Logstash are responsible for the comprehensive collection of log data; and data storage flexibly uses SQL or NoSQL databases to meet different needs. The data sources cover a wide range of dimensions, including base stations, core network equipment, and user terminals, ensuring wide coverage and accuracy of data. In addition, user complaint data records the reporting time, fault details, obstacle manifestations, and user location information in detail, providing solid support for troubleshooting and service optimization.

[0049] 2. External factor data integration: External factor data integration includes geographic information data, weather forecast data, and holiday and event information. Geographic information data obtains the location information and coverage of each base station through the Geographic Information System (GIS). Weather forecast data, such as temperature, humidity, and rainfall probability, can be obtained through the API interface provided by the Meteorological Bureau.

[0050] 3. For massive data, we use distributed database technologies such as Hadoop HDFS or Cassandra for storage; for data that needs to be accessed quickly, we use NoSQL databases such as Redis or MongoDB. In addition, we have formulated a data backup strategy, regularly backed up data, and ensured the ability to recover data to deal with possible data loss.

[0051] Step S204: construct a time series matrix for the complaint-related data in chronological order.

[0052] Optionally, constructing a time series matrix also includes preprocessing the data, which includes the following data processing flow: First, setting cleaning rules for the original data includes removing duplicate records, filling missing fields with reasonable default values, and identifying and processing obvious outliers to ensure the accuracy and completeness of the data. Next, a hierarchical weight system is set for the complaint data, where the first-level weight may be based on the importance of the complaint type, while the second-level weight may be further subdivided according to the specific content or scope of the complaint. Subsequently, a weighted average normalization method is used, with the number of complaints as the key feature, and the comprehensive score of each complaint is calculated in combination with the weights at all levels, so as to achieve standardized data processing and provide a unified metric for subsequent analysis.

[0053] In some technical solutions provided in step S204, constructing a time series matrix for complaint-related data in chronological order includes: preprocessing the complaint-related data; performing weighted normalization processing on the preprocessed complaint-related data; constructing a time series matrix for the weighted normalized complaint-related data in chronological order, wherein each row in the time series matrix corresponds to complaint-related data at a time point, and each column in the time series matrix corresponds to a type of complaint-related data.

[0054] Optionally, data preprocessing is a key step to ensure data quality. Data preprocessing mainly includes complaint data cleaning and weighted normalization.

[0055] Optionally, preprocessing the complaint-related data includes: determining a first data cleaning rule corresponding to the communication data in the complaint-related data, a second data cleaning rule corresponding to the customer complaint data, and a third data cleaning rule corresponding to the external data; cleaning the communication data according to the first data cleaning rule, cleaning the customer complaint data according to the second data cleaning rule, and cleaning the external data according to the third data cleaning rule.

[0056] Optionally, the first data cleaning rule includes deleting erroneous values, missing values ​​and extreme values ​​in the operating performance indicator data of the base station, and deleting the alarm information of the base station that is determined to not affect the business; the second data cleaning rule includes deleting non-target operator complaint data and non-network problem complaint data in the customer complaint data; the third data cleaning rule includes deleting missing values ​​in meteorological data.

[0057] Optionally, before data cleaning, determine the fields involved in the data type:

[0058]

[0059] Alternatively, let D = {d1, d2, ..., d n} is the original dataset, is a predicate used to determine whether the data item d i Should it be retained? Returns true, then the data item d i is retained; otherwise, data item d i will be deleted. A predicate P can be used uniformly, which contains the logical combination of all the removal conditions. In this way, the final cleaning result can be expressed as:

[0060]

[0061] (1) Base station KPI indicators

[0062] For the base station KPI indicators, the predicate P includes three parts: error value removal, extreme value removal, and missing value removal:

[0063]

[0064] (2) Base station alarm data

[0065] For base station alarm data, the predicate P determines whether the alarm affects the service:

[0066]

[0067] (3) Complaint data

[0068] For complaint data, predicate P includes two parts: excluding non-operator data and excluding non-network problem data:

[0069]

[0070] (4) Meteorological data

[0071] For meteorological data, the predicate P determines whether the data is missing:

[0072]

[0073] The meanings of the letters and symbols used in the above formula are:

[0074] 1.D: This is the symbolic representation of the raw data set, which represents the collection of all raw data being processed, including any type of data collection such as base station KPI indicators, base station alarm data, complaint data, or meteorological data.

[0075] 2.d i : Represents an element in the data set D, that is, a specific observation value or record.

[0076] 3.P di : is a predicate used to judge the data item d i Whether the retention condition is met. If P di Returns `true`, then d i will be retained; if it returns `false`, then d i Will be eliminated.

[0077] 4.D': This is the new data set after cleaning, which contains all data items that meet the retention conditions.

[0078] 5.ferr(k i ) For base station KPI indicator data, this is a function used to detect k i Is it an error value? i is valid, then return k i ; If it is an error value, a special marker is returned, such as `null` or `NaN`.

[0079] 6.u k : This is the average value of the effective value of the base station KPI indicator data set.

[0080] 7.o k : This is the standard deviation of the effective value of the base station KPI indicator data set.

[0081] 8.c: This is a constant that is usually used to define the range of outliers. For example, c = 3 usually means that any value that deviates from the mean by more than 3 standard deviations is considered an outlier.

[0082] 9.f biz(aj) : For base station alarm data, this is a function used to determine the alarm a j Whether it affects the business. If it does, it returns `true`; otherwise, it returns `false`.

[0083] 10.f op(ck) : For complaint data, this is a function used to determine the complaint c k

[0084] Whether it belongs to this operator. If yes, it returns `true`; otherwise, it returns `false`.

[0085] 11.f net(ck) : For complaint data, this is a function used to determine the complaint c k Whether network problems are involved. If yes, return `true`; otherwise return `false`.

[0086] 12.w l : For meteorological data, this is an element in the dataset W.

[0087] Through these definitions, we can define specific predicates P according to different data types and cleaning requirements. (di) , and perform data cleaning accordingly.

[0088] Optionally, weighted normalization processing is performed on the pre-processed complaint-related data, including: determining the first-level weight and second-level weight corresponding to each data in the pre-processed complaint-related data, wherein the first-level weight is used to reflect the importance of each type of data, and the second-level weight is used to reflect the scope of influence of the data; weighted normalization processing is performed on the pre-processed complaint-related data based on the first-level weight and the second-level weight.

[0089] Optionally, various types of indicator data are used as target variables, divided into primary weights and secondary weights according to the data type, and the various types of indicator data are normalized by hierarchical weighted average. The number of complaints is also input as a feature. Among them, the primary weight is used to reflect the importance of various types of data. In the data set, different data categories may have different degrees of influence on the final analysis results, and the primary weight is used to quantify this importance. The secondary weight is used to reflect the scope of influence of the data. In the same data category, different data items may have different contributions to the overall performance of the category, and the secondary weight is used to quantify this contribution. Optionally, for base station KPI indicators and meteorological data, when the base station KPI indicator has a greater impact on the final analysis results than the meteorological data, the primary weight of the base station KPI indicator is greater than the primary weight of the meteorological data indicator, for example, the primary weight of the base station KPI indicator is set to 0.5, and the primary weight of the meteorological data is set to 0.3. For KPI indicators such as wireless connection rate and switching success rate in the base station KPI indicators, when the influence range of the wireless connection rate is greater than the switching success rate, the secondary weight of the wireless connection rate indicator is greater than the secondary weight of the switching success rate indicator. For example, the secondary weight of the wireless connection rate is 0.3, and the secondary weight of the switching success rate is 0.2.

[0090] (1) Data average normalization

[0091]

[0092] Where x is a single data point in the original data set, mean(x) is the mean value of the original data set x, max(x) is the maximum value in the original data set, and min(x) is the minimum value in the original data set.

[0093] (2) Weighted average of each type of data

[0094]

[0095] n cat is the number of data items in the cat-th class. x,i is the weight of the i-th data item in the cat-th data (i.e., the secondary weight). i,cat is the value of the i-th data item.

[0096] (3) Final weighted average:

[0097]

[0098] Among them, W cat,j is the weight of the j-th class of data (i.e., the first weight), WA cat,j is the weighted average of the j-th category data.

[0099] The above three sub-formulas derive the weighted normalization formula as follows:

[0100]

[0101] Optionally, constructing the time series matrix includes arranging the normalized data in time order to form time series data. The data at each time point should contain all the selected features.

[0102] Assume that we have N observation time points, and at each time point t (where t = 1, 2, ..., N), we observe M features. The observed values ​​of these features at time point t can be represented as an M-dimensional vector xt, where xt = [xt(1), xt(2), ..., xt(M)]T.

[0103] The normalization process usually scales the value of each feature to a specific range (such as [0, 1] or [-1, 1]), which can be achieved through different normalization methods, such as min-max normalization, Z-score normalization, etc. Here, we assume that the normalized feature value is still represented by xt(i), where i is the index of the feature.

[0104] Time series data can be represented as a matrix X, where each row corresponds to all eigenvalues ​​at a time point. Therefore, the time series data matrix X can be defined as:

[0105]

[0106] The matrix X is the time series data containing all selected features in chronological order. Each row represents the observation value at a time point, and each column represents the observation value of a feature at different time points.

[0107] Step S206, using the customer behavior memory model to analyze the time series matrix to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area. The customer behavior memory prediction model is a prediction model used to analyze and predict time series data.

[0108] Optionally, the trained CBMM model is used to perform user complaint prediction analysis. Preferably, network status data is received in real time and input into the prediction model. The model will output prediction results, including the complaint risk level of each area in the future. Charts and reports are generated through data visualization tools to intuitively display the prediction results and key influencing factors. This helps operation and maintenance personnel to quickly understand the prediction results and take preventive measures accordingly. Using data visualization tools to generate charts and reports can not only help the operation and maintenance team understand the network status more intuitively, but also assist decision makers in formulating strategies to reduce future user complaints.

[0109] In some technical solutions provided in step S206, the customer behavior memory model includes an acceptance gate, an elimination gate, a storage core and an output regulator, wherein the acceptance gate is used to determine the data stored in the storage core and store the data; the elimination gate is used to determine the data to be deleted in the storage core and delete the data to be deleted; the storage core is used to generate candidate memory states based on the data stored in the storage core, wherein the candidate memory states are used to reflect the dependency relationship between the complaint behavior and the communication data and the external data; the output regulator is used to determine the target memory state from the candidate memory states, and output the prediction result based on the target memory state.

[0110] Optionally, Figure 4 A training process of CBMM is shown, such as Figure 4The CBMM model shown is trained by constructing a time series matrix. First, the normalized data is carefully constructed into a time series matrix in chronological order to ensure that each time slice includes key features. Then, a time series-based CBMM (Customer Behavior Memory Model) model is used for in-depth analysis. CBMM uses the acceptance gate to filter important information into the storage core, the elimination gate to remove invalid data, the storage core integrates information to generate memory states, and the output regulator accurately extracts the required information. This series of processed data is ultimately used as high-quality input to drive the trained CBMM model for efficient learning and prediction, providing strong support for optimizing services.

[0111] Optionally, the CBMM training adopts a set of CBMM (Customer Behavior Memory Model) models based on time series. First, the base station indicators, complaint volume and other data are used to construct a time series matrix. Then, the CBMM model is used for data processing, which includes four key steps:

[0112] (1) Elimination gate: determines which information should be deleted from the storage core.

[0113] f t =σ(W f *[h t-1 ,x t ]+b f )

[0114] f t : The output of the elimination gate determines the ratio of retention or forgetting.

[0115] σ: Sigmoid activation function, the output range is between 0 and 1.

[0116] W f : Weight matrix of the elimination gate.

[0117] [h t-1 ,x t ]: The concatenation vector of the hidden state at the previous time step and the input at the current time step.

[0118] b f : Bias term for the elimination gate.

[0119] (2) Admission gate: Determine which information should be stored in the storage core. The admission gate is divided into two parts: one is to determine which parts of the storage core should be updated; the other is to create a candidate storage core state c t .

[0120] The admission gate determines the Sigmoid layer that updates the storage kernel:

[0121] it =σ(W i *[h t-1 ,x t ]+b i )

[0122] i t : The output of the admission gate, which determines how many of the new candidate storage core states should be added to the cell state.

[0123] W i : Weight matrix of the admission gate.

[0124] b i : Bias term for the admission gate.

[0125] The admission gate creates a candidate storage core state:

[0126]

[0127] Candidates store nuclear states, representing new information that may be added to the cell state.

[0128] W c : The weight matrix of the candidate storage core state.

[0129] b c : Bias term for candidate storage core states.

[0130] tanh: Hyperbolic tangent activation function, the output range is between -1 and 1.

[0131] (3) Storage core: Stores the filtered information and generates candidate memory states, where the candidate memory state is a temporary representation of the information in the storage core, which reflects the current state of the information stored in the storage core. This state is generated based on the information filtered by the admission gate and the information retained by the elimination gate. The candidate memory state is used to reflect the dependency between the complaint behavior and the communication data and external data.

[0132]

[0133] c t : New cell state.

[0134] f t ⊙c t-1 :The result of applying the elimination gate by element-wise multiplication (Hadamard product) determines which information is removed from the old cell state c t-1 retained in.

[0135] The result of applying the admission gate via element-wise multiplication determines what information should be added to the cell state.

[0136] ⊙: represents element-by-element multiplication.

[0137] (4) Output regulator: Determines which information should be output. Specifically, the output regulator selects the most relevant information from the candidate memory states based on the current prediction task and the learning goal of the model.

[0138] o t =σ(W o *[h t-1 ,x t ]+b o )

[0139] o t : The output of the output regulator, which determines how much information in the cell state should be output as the hidden state of the current time step.

[0140] W o : The weight matrix of the output controller.

[0141] b o : Bias term for the output regulator.

[0142] The hidden state is the result of storing the kernel state after Tanh activation and multiplying it with the output regulator.

[0143] h t =o t ⊙tanh (c t )

[0144] h t : The hidden state of the current time step, which is passed to the next time step as part of the input and usually also serves as the output of that time step.

[0145] tanh(c t ): A hyperbolic tangent activation function is applied to the cell state to ensure that the output is between -1 and 1.

[0146] o t ⊙tanh(c t ): The hidden state of the output is determined by applying the result of the output regulator through element-wise multiplication.

[0147] Finally, the processed data is fed into the trained model.

[0148] Optionally, after analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, the method further includes: generating a prediction display chart based on the prediction result, and displaying the prediction display chart.

[0149] Optionally, in the prediction process, the first step is real-time data input, that is, inputting the data captured from various channels in real time into the trained prediction model. Then, the model will output the complaint risk level of each area in the future period based on these data.

[0150] Optionally, in order to better understand and communicate the forecast results, we use professional tools such as Matplotlib, Seaborn or Tableau to generate charts and reports. Through these tools, the forecast results and their key influencing factors can be presented in the form of bar charts, heat maps and other charts, making the information more intuitive and easy to understand. Figure 5 An example of complaint prediction results is shown.

[0151] Through the above steps, a user complaint prediction method based on the CBMM model can be implemented, including collecting and integrating multi-source information such as base station operation data, user complaint data and meteorological data, and performing refined cleaning and weighted normalization processing, using a time series-based CBMM model to construct a time series matrix using preprocessed data, and training the CBMM model based on the matrix to predict and analyze user behavior and network status. First, through advanced prediction algorithms, the embodiments of the present application can provide insight into and warn of potential network problems in advance, implement proactive prevention strategies, and thus significantly reduce the occurrence rate and scope of failures. Secondly, in response to user complaints, the embodiments of the present application use refined data analysis and processing technology, which not only effectively reduces the cost of complaint handling, but also significantly reduces the number of complaints faced by operators, and improves service efficiency and brand image.

[0152] Furthermore, by intelligently managing base station operations, the embodiments of the present application achieve optimal allocation and utilization of resources, effectively reduce operating costs, and bring long-term economic benefits to operators. The ultimate goal of all these efforts is to improve user satisfaction and experience. By responding to user needs in real time and optimizing network service quality, the embodiments of the present application ensure that users can feel the care and respect from operators while enjoying high-speed and stable communication services.

[0153] In summary, the embodiments of the present application are based on CBMM technology and are committed to solving various problems in communication networks, achieving a comprehensive leap in cost savings, efficiency improvement, and user satisfaction.

[0154] In view of this, the patent of this invention innovatively integrates multi-source information such as base station network management, complaint system and weather data API to ensure the timeliness and accuracy of the data. The data quality is significantly improved through refined data cleaning and weighted normalization technology. Furthermore, the Customer Behavior Memory Model (CBMM) is trained using the constructed time series matrix to achieve instant prediction and in-depth analysis of user complaints. This move not only strengthens the operator's control over service quality, but also more effectively improves user satisfaction, laying a solid foundation for building a better communication environment.

[0155] The present application embodiment provides a complaint prediction device, Figure 6 is a schematic diagram of the structure of the device, such as Figure 6 As shown, the device includes: a first processing module 60, which is used to obtain complaint-related data of a target area, wherein the complaint-related data includes communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of the base station, and the external data includes meteorological data of the target area; a second processing module 62, which is used to construct a time series matrix for the complaint-related data in chronological order; a third processing module 64, which is used to analyze the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model used to analyze and predict time series data.

[0156] In some embodiments of the present application, the customer behavior memory model includes an acceptance gate, an elimination gate, a storage core and an output regulator, wherein the acceptance gate is used to determine the data stored in the storage core and store the data; the elimination gate is used to determine the data to be deleted in the storage core and delete the data to be deleted; the storage core is used to generate candidate memory states based on the data stored in the storage core, wherein the candidate memory states are used to reflect the dependency relationship between the complaint behavior and the communication data and the external data; the output regulator is used to determine the target memory state from the candidate memory states, and output the prediction result based on the target memory state.

[0157] In some embodiments of the present application, constructing a time series matrix for complaint-related data in chronological order includes: preprocessing the complaint-related data; performing weighted normalization processing on the preprocessed complaint-related data; and constructing a time series matrix for the weighted normalized complaint-related data in chronological order, wherein each row in the time series matrix corresponds to complaint-related data at a time point, and each column in the time series matrix corresponds to a type of complaint-related data.

[0158] In some embodiments of the present application, preprocessing of complaint-related data includes: determining a first data cleaning rule corresponding to communication data in the complaint-related data, a second data cleaning rule corresponding to customer complaint data, and a third data cleaning rule corresponding to external data; cleaning the communication data according to the first data cleaning rule, cleaning the customer complaint data according to the second data cleaning rule, and cleaning the external data according to the third data cleaning rule.

[0159] In some embodiments of the present application, the first data cleaning rule includes deleting error values, missing values ​​and extreme values ​​in the operating performance indicator data of the base station, and deleting the alarm information of the base station that is determined to not affect the business; the second data cleaning rule includes deleting non-target operator complaint data and non-network problem complaint data in the customer complaint data; the third data cleaning rule includes deleting missing values ​​in meteorological data.

[0160] In some embodiments of the present application, weighted normalization processing is performed on the pre-processed complaint-related data, including: determining the primary weight and secondary weight corresponding to each data in the pre-processed complaint-related data, wherein the primary weight is used to reflect the importance of the data, and the secondary weight is used to reflect the scope of influence of the data; weighted normalization processing is performed on the pre-processed complaint-related data according to the primary weight and the secondary weight.

[0161] In some embodiments of the present application, after analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, the method further includes: generating a prediction display chart based on the prediction result, and displaying the prediction display chart.

[0162] It should be noted that the various modules in the above-mentioned complaint prediction device can be program modules (for example, a set of program instructions that implement a certain specific function) or hardware modules. For the latter, it can be expressed in the following forms, but is not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.

[0163] An embodiment of the present application provides a non-volatile storage medium, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the following complaint prediction method: obtaining complaint-related data of a target area, the complaint-related data including communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of a base station, and the external data includes meteorological data of the target area; constructing a time series matrix for the complaint-related data in chronological order; using a customer behavior memory model to analyze the time series matrix to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model for analyzing and predicting time series data.

[0164] An embodiment of the present application provides an electronic device, including: a memory and a processor, the processor being used to run a program stored in the memory, wherein the following complaint prediction method is executed when the program is running: obtaining complaint-related data of a target area, the complaint-related data including communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of a base station, and the external data includes meteorological data of the target area; constructing a time series matrix for the complaint-related data in chronological order; and analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes a complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model for analyzing and predicting time series data.

[0165] An embodiment of the present application provides a computer program product, including a computer program, which implements the following complaint prediction method when executed by a processor: obtaining complaint-related data of a target area, the complaint-related data including communication data, customer complaint data and external data, wherein the communication data includes operating performance indicator data and real-time alarm information of a base station, and the external data includes meteorological data of the target area; constructing a time series matrix for the complaint-related data in chronological order; and analyzing the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes a complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model for analyzing and predicting time series data.

[0166] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0167] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0168] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0169] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0170] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the relevant technology or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0171] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A complaint prediction method, characterized in that: include: Acquire complaint-related data of the target area, wherein the complaint-related data includes communication data, customer complaint data and external data, wherein the communication data includes operating performance index data and real-time alarm information of the base station, and the external data includes meteorological data of the target area; Constructing a time series matrix for the complaint-related data in chronological order; The time series matrix is ​​analyzed using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area. The customer behavior memory prediction model is a prediction model used to analyze and predict time series data.

2. The complaint prediction method according to claim 1, characterized in that: The customer behavior memory model includes an acceptance gate, an elimination gate, a storage core and an output regulator, wherein: The admission gate is used to determine the data to be stored in the storage core and store the data; The elimination gate is used to determine the data to be deleted in the storage core and delete the data to be deleted; The storage core is used to generate a candidate memory state according to the data stored in the storage core, wherein the candidate memory state is used to reflect the dependency relationship between the complaint behavior and the communication data and the external data; The output regulator is used to determine a target memory state from the candidate memory states and output the prediction result according to the target memory state.

3. The complaint prediction method according to claim 1, characterized in that: The time series matrix constructed according to the time sequence of the complaint-related data includes: Preprocessing the complaint-related data; Performing weighted normalization processing on the pre-processed complaint-related data; The complaint-related data after weighted normalization processing is constructed into a time series matrix in chronological order, wherein each row in the time series matrix corresponds to complaint-related data at a time point, and each column in the time series matrix corresponds to a type of complaint-related data.

4. The complaint prediction method according to claim 3, characterized in that: Preprocessing the complaint-related data includes: Determine a first data cleaning rule corresponding to the communication data in the complaint-related data, a second data cleaning rule corresponding to the customer complaint data, and a third data cleaning rule corresponding to the external data; The communication data is cleaned according to the first data cleaning rule, the customer complaint data is cleaned according to the second data cleaning rule, and the external data is cleaned according to the third data rule.

5. The complaint prediction method according to claim 4, characterized in that: The first data cleaning rule includes deleting error values, missing values ​​and extreme values ​​in the operation performance indicator data of the base station, and deleting alarm information of the base station that is determined to have no impact on the service; The second data cleaning rule includes deleting non-target operator complaint data and non-network problem complaint data in the customer complaint data; The third data cleaning rule includes deleting missing values ​​in the meteorological data.

6. The complaint prediction method according to claim 3, characterized in that: The weighted normalization processing of the pre-processed complaint-related data includes: Determine the primary weight and secondary weight corresponding to each data in the pre-processed complaint-related data, wherein the primary weight is used to reflect the importance of the data, and the secondary weight is used to reflect the impact scope of the data; The pre-processed complaint-related data is weighted and normalized according to the primary weight and the secondary weight.

7. The complaint prediction method according to claim 1, characterized in that: After analyzing the time series matrix using the customer behavior memory model to obtain a prediction result, the method further includes: Generate a prediction display chart based on the prediction result, and display the prediction display chart.

8. A complaint prediction device, characterized in that: include: A first processing module is used to obtain complaint-related data of a target area, wherein the complaint-related data includes communication data, customer complaint data and external data, wherein the communication data includes operation performance index data and real-time alarm information of a base station, and the external data includes meteorological data of the target area; The second processing module is used to construct a time series matrix for the complaint-related data in chronological order; The third processing module is used to analyze the time series matrix using a customer behavior memory model to obtain a prediction result, wherein the prediction result includes the complaint risk level of the target area, and the customer behavior memory prediction model is a prediction model used to analyze and predict time series data.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the complaint prediction method described in any one of claims 1 to 7.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program executes the complaint prediction method described in any one of claims 1 to 7 when running.

11. A computer program product, characterized in that It comprises a computer program which, when executed by a processor, implements the complaint prediction method according to any one of claims 1 to 7.

Images