User satisfaction prediction method and device, equipment, medium and product

Abstract

The invention relates to the technical field of network maintenance, in particular to a user satisfaction prediction method and device, equipment, a medium and a product, and the method comprises the steps: obtaining fusion data affecting the user satisfaction; calculating index data corresponding to each quantitative index based on a plurality of preset quantitative indexes and the fusion data; calculating a proportionality coefficient of each quantitative index based on the fluctuation degree of the index data; and obtaining a calculation formula for predicting the user satisfaction according to each quantitative index and the proportionality coefficient. A user portrait is drawn through call voice and an operator system, the index data corresponding to each quantitative index is calculated, the proportionality coefficient of each quantitative index is determined through the fluctuation degree of the index data, a calculation formula of the user satisfaction can be clearly obtained, the user satisfaction can be objectively and accurately predicted, and the user experience is improved. And the flexibility and the accuracy of the service of the operator can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of network maintenance technology, and in particular to a method, apparatus, device, medium and product for predicting user satisfaction. Background Technology

[0002] With the deep coverage of 5G networks and the large-scale access of IoT terminals, the diversity of user network behavior patterns and the real-time requirements of business scenarios are growing exponentially. Therefore, in order to provide users with more accurate and suitable services, it is necessary to accurately predict user satisfaction and its evolution trend. For example, satisfaction prediction technology has been deeply integrated into fields such as e-commerce, healthcare, and tourism.

[0003] However, in the field of operator services, there is still no clear method for predicting satisfaction. The current management and care mechanism for user satisfaction mainly relies on standardized templates, which leads to a lack of flexibility and accuracy in operator services and makes it impossible to effectively improve user perception. Summary of the Invention

[0004] In view of the above problems, this disclosure is made to provide a method, apparatus, device, medium and product for predicting user satisfaction.

[0005] According to one aspect of this disclosure, a method for predicting user satisfaction is provided, comprising: acquiring fused data that affects user satisfaction, wherein the fused data is obtained based on voice call data and performance data of the operator system; Based on multiple preset quantitative indicators and integrated data, the corresponding indicator data for each quantitative indicator is calculated; among them, the quantitative indicators are used to characterize the user's tolerance, scenario needs and emotional intensity. Calculate the proportional coefficients of each quantitative indicator based on the degree of fluctuation in the indicator data; Based on various quantitative indicators and proportional coefficients, a calculation formula for predicting user satisfaction is obtained.

[0006] The technical benefits of this solution are as follows: by drawing user profiles through voice calls and operator systems, the corresponding indicator data for each quantitative indicator is calculated, and the proportional coefficient of each quantitative indicator is determined by the fluctuation of the indicator data. This allows for a clear formula for calculating user satisfaction, enabling objective and accurate prediction of user satisfaction, and improving the flexibility and accuracy of operator services.

[0007] In addition, according to the user satisfaction prediction method of one aspect of this disclosure, the quantitative indicators include fluctuation tolerance, scenario sensitivity coefficient, service timeliness demand duration, voice emotion recognition coefficient and text difference coefficient. Based on multiple preset quantitative indicators and fused data, the corresponding indicator data for each quantitative indicator is calculated, including: The network performance data in the fused data is obtained, and the index data corresponding to the fluctuation tolerance is calculated based on the pre-trained neural network model and the network performance data. Among them, the network performance data includes packet loss rate, latency and jitter, and the fluctuation tolerance is used to predict the probability of user-initiated complaints. In addition, traffic data is acquired from the fused data, and based on the traffic data, the corresponding indicator data of the scenario sensitivity coefficient is calculated; where traffic data is the data generated when users access application software, and the scenario sensitivity coefficient is used to characterize the user's scenario needs. Furthermore, based on the fused data, identify the user's identity characteristics; based on the identity characteristics, calculate the indicator data corresponding to the service timeliness demand duration; Additionally, the system acquires speech feature data from the fused data, and calculates the corresponding index data for speech emotion recognition coefficient and text difference coefficient based on the speech feature data. The speech feature data is used to characterize the user's emotions and the intensity of their word choice.

[0008] In one or more embodiments, the technical effect of this solution is as follows: by setting the quantitative indicators as fluctuation tolerance, scenario sensitivity coefficient, service timeliness requirement duration, voice emotion recognition coefficient, and text difference coefficient, it covers users' tolerance for network fluctuations; users' scenario needs, such as whether they like playing games (afraid of latency) or watching videos (afraid of buffering); users' patience, such as business users may be less patient than student users; and the emotional tone and intensity of users' speech. This facilitates the creation of targeted user profiles and enables the objective and accurate calculation of the proportional coefficients of each quantitative indicator.

[0009] Furthermore, according to one aspect of the user satisfaction prediction method disclosed herein, based on the degree of fluctuation of indicator data, the proportional coefficients of each quantitative indicator are calculated, including: The indicator data is subjected to min-max normalization to obtain the normalized data corresponding to each quantitative indicator; Based on normalized data, the information entropy of each quantitative indicator is calculated; whereby information entropy is used to characterize the degree of fluctuation of normalized data. Calculate the information utility value of each quantitative indicator based on information entropy; Calculate the sum of information utility values, divide the information utility value of each quantitative indicator by the sum, and obtain the proportional coefficient of each quantitative indicator.

[0010] In one or more embodiments, the technical effect of this solution is that it calculates the proportional coefficient of each quantitative indicator using the entropy weight method, i.e., what proportion each quantitative indicator should account for in the calculation formula. This solves the problem that there is no clear calculation formula in the current operator service field, making user satisfaction calculable and predictable.

[0011] Furthermore, according to a user satisfaction prediction method of one aspect of this disclosure, fused data affecting user satisfaction is obtained, including: The performance data is obtained from the operator's system, including service ticket structured data and network performance data. If a phone conversation has been conducted with the user, obtain the user's voice recordings. The voice call is analyzed and processed to extract voice feature data. The voice feature data includes voice text data, emotional state data, and entity label data. The entity label data is obtained based on the core needs of the voice call. By fusing performance data and voice feature data, we can obtain fused data that influences user satisfaction.

[0012] In one or more embodiments, the technical effect of this solution is that: acquiring data from multiple data sources such as operator systems and voice calls helps to create accurate user profiles and improve the accuracy of predictions.

[0013] Furthermore, according to one aspect of the user satisfaction prediction method of this disclosure, before calculating the indicator data corresponding to each quantitative indicator based on multiple preset quantitative indicators and fused data, the method further includes: By utilizing graph databases and fused data, a relationship network is established, which is used to represent the ternary relationship between users, the services enjoyed by users, and the networks used by users. Based on the relationship network, it was confirmed that the fused data did not reach a specific threshold; where the specific threshold is represented by the maximum probability threshold of user-initiated complaints.

[0014] In one or more embodiments, the technical effect of this solution is as follows: by establishing a relationship network through a graph database, the connection between the user, the service enjoyed by the user, and the network used by the user can be seen intuitively, which is conducive to performing relationship reasoning and quickly determining whether the user has reached a specific threshold. For example, when a user experiences a delay of >500ms three times in a row, it can be considered that a specific threshold has been reached, that is, the user is very likely to file a complaint. In this case, the user should be cared for in a timely manner to improve user satisfaction.

[0015] Furthermore, the user satisfaction prediction method according to one aspect of this disclosure also includes: If user satisfaction is less than the preset threshold, a care strategy for the user will be generated based on the user's data under various quantitative indicators. And, based on the calculation formula, predict the user satisfaction of all users; Based on the user satisfaction of all users, target users are selected, including at least one of the following: potential churned users and high-value users.

[0016] In one or more embodiments, the technical effect of this solution is as follows: by determining whether user satisfaction is less than a preset threshold, proactive warnings can be issued before user complaints are received. Care strategies can include providing game users with speed-up packages, video users with high-definition data packages, or notifying users who frequently visit the area before network maintenance. By identifying target users, such as discovering a user with extremely low data usage or consistently exceeded bandwidth limits for a week (potential churned user), the target user can be proactively contacted and maintained by an account manager.

[0017] According to another aspect of this disclosure, a user satisfaction prediction device is provided, comprising: The acquisition module is used to acquire fused data that affects user satisfaction. The fused data is obtained based on voice call data and performance data of the operator's system. The first calculation module is used to calculate the indicator data corresponding to each quantitative indicator based on multiple preset quantitative indicators and fused data; among them, the quantitative indicators are used to characterize the user's tolerance, scenario needs and emotional intensity. The second calculation module is used to calculate the proportional coefficient of each quantitative indicator based on the degree of fluctuation of the indicator data. The formula generation module is used to obtain calculation formulas for predicting user satisfaction based on various quantitative indicators and proportional coefficients.

[0018] According to another aspect of this disclosure, a computer device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method of one aspect above.

[0019] According to another aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method of one aspect above.

[0020] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method of the above-described aspect.

[0021] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0022] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 This is a system architecture diagram illustrating a user satisfaction prediction method according to an embodiment of this disclosure.

[0024] Figure 2 This is a flowchart illustrating the calculation method of the scaling factor A according to an embodiment of the present disclosure.

[0025] Figure 3 This is a flowchart illustrating a user satisfaction prediction method according to an embodiment of this disclosure.

[0026] Figure 4 This is a schematic diagram illustrating the structure of a user satisfaction prediction device according to an embodiment of the present disclosure.

[0027] Figure 5 This is a schematic diagram illustrating the structure of a computer device according to an embodiment of the present disclosure.

[0028] Figure 6 This is a schematic diagram illustrating a computer program product according to an embodiment of the present disclosure. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this disclosure more apparent, exemplary embodiments according to this disclosure will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments of this disclosure. It should be understood that this disclosure is not limited to the exemplary embodiments described herein.

[0030] With the deep coverage of 5G networks and the large-scale access of IoT terminals, the diversity of user network behavior patterns and the real-time requirements of business scenarios are growing exponentially. Therefore, in order to provide users with more accurate and suitable services, it is necessary to accurately predict user satisfaction and its evolution trend. For example, satisfaction prediction technology has been deeply integrated into fields such as e-commerce, healthcare, and tourism.

[0031] However, in the field of operator services, there is still no clearly defined method for predicting customer satisfaction. Operators are involved in data-sensitive activities, and their approach differs from other sectors. The operator service sector does not fully utilize data from user interactions, such as follow-up visits and complaint records. There is no clear formula for predicting operator satisfaction; current customer satisfaction management and care mechanisms mainly rely on standardized templates. In network failure scenarios, a uniform script is often used for proactive care, resulting in a lack of flexibility and accuracy in service delivery. This passive approach fails to effectively improve customer perception and is not applicable to operator services.

[0032] The above description, with reference to the accompanying drawings, illustrates a user satisfaction prediction method, apparatus, device, medium, and product according to embodiments of the present disclosure. By drawing user profiles through voice calls and operator systems, calculating the corresponding indicator data for each quantitative indicator, and determining the proportional coefficient of each quantitative indicator based on the fluctuation of the indicator data, a clear formula for calculating user satisfaction can be obtained. This allows for objective and accurate prediction of user satisfaction, which is beneficial for improving the flexibility and accuracy of operator services.

[0033] To facilitate understanding of this embodiment, a user satisfaction prediction method disclosed in this disclosure will first be described in detail. The execution entity of the user satisfaction prediction method provided in this disclosure is generally a computer device with certain computing capabilities. This computer device may include, for example, a terminal device, a server, or other processing devices. The terminal device may be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, this user satisfaction prediction method can be implemented by a processor calling computer-readable instructions stored in memory.

[0034] like Figure 1 The diagram shown is a system architecture diagram of the user satisfaction prediction method provided in this embodiment of the disclosure, including a telephone analysis module 1, a user profile generation module 2, a proportional coefficient calculation module 3, and a repair module 4. The specific functional flow of each module is described below: Telephone Analysis Module 1: Analyzes user voice recordings after telephone conversations and extracts voice feature data.

[0035] Voice feature data includes voice-text data, entity label data, and emotional state data. Specifically, an AI voice analysis engine and neural network model are used to extract voiceprint features from call speech to capture the user's emotional state data. A semantic parsing framework is built based on the BERT-wwm pre-trained model to extract the content of the call speech into voice-text data. Then, dependency analysis is used to extract the core keywords of the call speech and obtain entity label data, such as automatically labeling entity labels like "network latency > 500ms" and "package tariff dispute".

[0036] The study employed a suitable AI voice analysis engine and a deep neural network (Wave2Vec 2.0) to extract voiceprint features from call recordings, and combined this with the Open SMILE toolkit to identify 137 acoustic parameters, including fundamental frequency, speech rate, and pause intervals. A Bi-LSTM emotion classification model was used to capture real-time data on seven emotional states, including anger (e.g., high-frequency sound intensity) and anxiety (e.g., irregular breathing sounds). Experimental results demonstrated an accuracy of 92% in emotion classification.

[0037] It should be noted that if the user has not made a phone complaint before, this module can be skipped, and only the performance data obtained from the operator's system should be referenced when creating the user profile.

[0038] User profile generation module 2: Specifically includes the following functions: (1) Data fusion The system collects and organizes performance data from the operator's system, specifically from the government and enterprise operation and maintenance system, including structured service work order data such as processing time and problem classification; and from the network management and performance system, including network performance data such as packet loss rate and latency fluctuation.

[0039] The voice feature data extracted by the telephone analysis module 1 is fused with the service work order structured data and network performance data to obtain fused data.

[0040] (2) Constructing a knowledge graph Using the Neo4j graph database and fused data, a "user-service-network" relationship network is established to represent the ternary relationship between users, the services they enjoy, and the networks they use. Core entities include: User node: Basic attributes (package type / terminal device); Service node: Work order processing records (response time / resolution rate); Network node: Base station location / historical fault events; By using GraphSAGE graph neural networks to perform relationship reasoning on the relationship network, when the fused data (or relationship network) reaches a certain threshold, proactive care is immediately provided to the user. For example, if a user experiences a delay of more than 500ms three times consecutively, it can be considered that a certain threshold has been reached. There is no need to predict the user's satisfaction anymore; proactive care and appropriate remedial measures should be taken immediately. Conversely, if the certain threshold is not reached, the process proceeds to the next part: feature engineering.

[0041] (3) Feature Engineering Based on the obtained voice and text data, entity label data, emotional state data, processing time, problem classification, packet loss rate, latency fluctuation and other data, the specific indicator data of each quantitative indicator is calculated based on multiple preset quantitative indicators.

[0042] The quantitative indicators include fluctuation tolerance, scenario sensitivity coefficient, service timeliness requirement duration, voice emotion recognition coefficient, and text difference coefficient. The quantification methods for each indicator are as follows: Fluctuation Tolerance: Based on a pre-trained neural network model and network performance data such as packet loss rate and latency fluctuations, the fluctuation tolerance metric is calculated. The neural network model uses a time-series prediction approach, is applied to all users, and a new model is trained every two days. The input is network performance data from the past 72 hours, and the output is the probability of predicting user-initiated complaints, which is the fluctuation tolerance metric.

[0043] Scene Sensitivity Coefficient: Traffic data is acquired from the fused data set. Based on this traffic data, the corresponding indicator data for the scene sensitivity coefficient is calculated. Specifically, consistent with the time interval of the neural network model used for fluctuation tolerance, traffic data of user access to the application software is obtained every two days based on traffic data detection technology. Users are differentiated and tagged according to the traffic data, and differences in user business scenarios are identified through an attention mechanism. Combined with survey and literature data, the indicator data corresponding to the scene sensitivity coefficient is obtained. The scene sensitivity coefficient is used to characterize the user's scenario requirements. For example, if a user is playing a game, the scenario requirement should be no latency; if a user is watching a video, the scenario requirement should be no lag.

[0044] Service timeliness demand duration: Based on integrated data, user identity characteristics are identified; based on these characteristics, corresponding indicator data for service timeliness demand duration are calculated. Specifically, a survival analysis model (Cox proportional hazards model) can be used to calculate the expected resolution time for users with different identities; for example, the acceptable time for business users is shorter than that for students.

[0045] Voice emotion recognition coefficient: Based on emotion state data, calculate the index data corresponding to the voice emotion recognition coefficient, and combine it with the BiGRU network to realize 7 types of emotion classification (anger / anxiety / satisfaction, etc.).

[0046] Text Difference Coefficient: Based on speech text data, entity label data, and emotional state data, the index data corresponding to the text difference coefficient is calculated to identify high and low emotional intensity features, as shown in Table 1, which is a quantitative analysis table of text difference coefficient. Table 1. Quantitative Analysis of Text Difference Coefficients

[0047] Example calculation: Assuming the functional impact level has "lag" (lag = 4) and "local impact" (local impact = 0.5), the time range description has "occasional" (occasional = 0.3), and the punctuation intensity has "exclamation mark" (exclamation mark = 0.5), then the text difference coefficient = (4 + 0.5). 0.675.

[0048] Finally, based on the integrated data and the corresponding indicator data of each quantitative indicator, a targeted user profile is created.

[0049] Proportion coefficient calculation module 3: Used to calculate the proportion coefficient of each quantitative indicator.

[0050] The formula for calculating user satisfaction is: User satisfaction = A Fluctuation tolerance +B Scene sensitivity coefficient +C Service timeliness requirement duration + D Voice emotion recognition coefficient +E Textual difference coefficient.

[0051] Therefore, it is necessary to calculate the proportionality coefficients A, B, C, D, and E. Based on the corresponding indicator data for each quantitative indicator, and combined with historical data, each proportionality coefficient (i.e., entropy weight) is calculated. Information entropy is an indicator in information theory that measures the uncertainty of a system; its calculation formula is:

[0052] Where X represents a random variable, and P(x) represents the probability of event x occurring. The unit of information entropy is usually bit. This embodiment uses fluctuation tolerance as an example, such as... Figure 2 The diagram shown is a flowchart of the calculation method for the proportionality coefficient A, including S201-S204: S201: Minimum-Maximum Normalization Process: The indicator data and historical data are subjected to Min-Max normalization to eliminate the influence of units, resulting in normalized data. The calculation formula is as follows:

[0053] in, This represents the i-th normalized data point. This represents the i-th original data (including indicator data and historical data). This represents historical data.

[0054] Table 2 shows an example table of normalized data: Table 2 Example table of normalized data

[0055] S202: Calculate the probability distribution: Calculate the probability of each sample for the normalized data:

[0056] in, Represents probability. This represents the k-th normalized data point. This is used to avoid zero values ​​(Laplace smoothing).

[0057] For example, suppose the data is normalized. =0.83, the normalized sum of the total samples is 125478.3, then the probability is:

[0058] S203: Calculate information entropy: Specifically, the information entropy is calculated item by item using the formula for calculating information entropy mentioned above:

[0059] Calculate ln500000 ,

[0060] in, Information entropy is represented and calculated for each sample. ,For example: when , ,

[0061] Accumulate all samples Assuming the cumulative sum is -9.44, then the final information entropy is:

[0062] That is, the information entropy of the volatility tolerance is 0.72. Following the above procedure, the information entropy of each quantitative indicator is calculated sequentially.

[0063] S204: Calculate the information utility value and then calculate the proportionality coefficient based on the information utility value.

[0064] Information utility value That is, the information utility value of the volatility tolerance is 0.28. That is, the proportionality coefficient of the fluctuation tolerance is 0.3.

[0065] Table 3 shows an example table of the proportional coefficients for various quantitative indicators: Table 3. Example Table of Proportion Coefficients for Quantitative Indicators

[0066] Based on Table 3, the formula for calculating user satisfaction can be obtained as follows: User satisfaction = 0.3 Fluctuation tolerance +0.21 Scene sensitivity coefficient +0.35 Service delivery time required +0.09 Voice emotion recognition coefficient +0.05 Textual difference coefficient.

[0067] Based on the above calculation formula, continuous monitoring of the user network is implemented. Once the network status meets the preset triggering rules below, a passive early warning is triggered, and a list of users requiring attention is pushed to the business side. When network fluctuations affect user experience, user perception will be predicted based on the satisfaction coefficient. If it exceeds the threshold of this user group, proactive care will be initiated.

[0068] Repair Module 4: Includes active and passive early warning systems. (1) Proactive early warning: Triggering mechanism: When the predictive model detects a user satisfaction level < 0.7, a corresponding care strategy is generated. Note that 0.7 is only an example threshold; the threshold varies depending on the user type. For example, it is 0.9 for 3A key users, 0.8 for government and enterprise users, and 0.7 for ordinary users.

[0069] Care strategies include: Real-time network compensation: For example, generating differentiated data packages based on user plan type (such as pushing exclusive acceleration packages to game users, and giving 4K ultra-high-definition data to video users).

[0070] Predictive notifications: For example, in conjunction with the data center maintenance plan, send a customized reminder 3 days in advance: "The network in your frequently used community will be optimized from 23:00 to 6:00 tomorrow night. It is recommended to cache videos in advance."

[0071] Targeted care generation: By combining the user profile, historical care cases, and the aforementioned care strategies, a targeted care plan is generated using large-scale modeling technology, thereby improving the flexibility and accuracy of proactive care.

[0072] (2) Passive early warning: While performing traffic and performance checks on all users, if the rules shown in Table 4 are triggered, the user is marked as a potential churner or a high-value user, and the alert is pushed to the front end for proactive maintenance by the front end manager. Table 4 shows the rules for passive alerts. Table 4. Rules for Passive Early Warning

[0073] Based on the above embodiments, this embodiment also provides a user satisfaction prediction method, such as... Figure 3 The flowchart shown is for the user satisfaction prediction method, including S301-S304: S301: Obtain fused data that affects user satisfaction.

[0074] The fused data is obtained based on voice call data and performance data from the operator's system.

[0075] S302: Based on multiple preset quantitative indicators and fused data, calculate the indicator data corresponding to each quantitative indicator.

[0076] Among them, quantitative indicators are used to characterize users' tolerance, scenario needs, and the intensity of their emotions.

[0077] S303: Calculate the proportional coefficient of each quantitative indicator based on the degree of fluctuation of the indicator data.

[0078] S304: Based on the various quantitative indicators and proportional coefficients, obtain the calculation formula for predicting user satisfaction.

[0079] Specifically, the quantitative indicators include fluctuation tolerance, scenario sensitivity coefficient, service timeliness requirement duration, voice emotion recognition coefficient, and text difference coefficient. Among them, fluctuation tolerance is used to predict the probability of users actively complaining, scenario sensitivity coefficient is used to characterize users' scenario needs, service timeliness requirement duration is used to characterize users' patience, voice emotion recognition coefficient is used to characterize the intensity of users' emotions, and text difference coefficient is used to characterize users' emotional state.

[0080] In one or more embodiments, based on a plurality of preset quantitative indicators and fused data, indicator data corresponding to each quantitative indicator is calculated, including: Fluctuation Tolerance: This involves acquiring network performance data from the fused data set. Based on a pre-trained neural network model and the network performance data, the corresponding metric for fluctuation tolerance is calculated. Network performance data includes packet loss rate, latency, and jitter. The neural network model uses a time-series prediction approach, is trained on a new model every two days for all users. Inputting network performance data from the past 72 hours, the output is the probability of predicting user-initiated complaints, which constitutes the fluctuation tolerance metric.

[0081] Scene Sensitivity Coefficient: This involves acquiring traffic data from the fused data set and calculating the corresponding indicator data for the scene sensitivity coefficient. Traffic data refers to data generated when users access the application software. Specifically, consistent with the time interval of the neural network model used for fluctuation tolerance, traffic data from user access to the application software is obtained every two days based on traffic data detection technology. Users are differentiated and tagged according to the traffic data, and differences in user business scenarios are identified through an attention mechanism. Combined with survey and literature data, the indicator data corresponding to the scene sensitivity coefficient is obtained. The scene sensitivity coefficient characterizes the user's scenario requirements; for example, if a user is playing a game, the scenario requirement should be no latency; if a user is watching a video, the scenario requirement should be no lag.

[0082] Service timeliness demand duration: Based on fused data, identify user identity characteristics; based on identity characteristics, calculate the indicator data corresponding to the service timeliness demand duration; among them, the survival analysis model (Cox proportional hazards model) can be used to calculate the expected resolution time for users with different identities, such as the acceptable time for business users is shorter than that for students.

[0083] Voice emotion recognition coefficient: Acquire voice feature data from the fused data, and calculate the corresponding index data of voice emotion recognition coefficient based on the voice feature data. The voice feature data is used to characterize the user's emotions and the intensity of the words used. The voice feature data includes emotional state data.

[0084] Text Difference Coefficient: Based on speech feature data, the index data corresponding to the text difference coefficient is calculated to identify high and low emotional intensity features. Speech feature data includes speech-text data, entity label data, and emotional state data. Table 1 above shows the quantitative analysis table for the text difference coefficient.

[0085] In one or more embodiments, the proportional coefficient of each quantitative indicator is calculated based on the volatility of the indicator data, including: The indicator data is subjected to min-max normalization to obtain normalized data for each quantitative indicator; based on the normalized data, the information entropy of each quantitative indicator is calculated; whereby the information entropy is used to characterize the degree of fluctuation of the normalized data; based on the information entropy, the information utility value of each quantitative indicator is calculated; the sum of the information utility values ​​is calculated, and the information utility value of each quantitative indicator is divided by the sum to obtain the proportional coefficient of each quantitative indicator.

[0086] Specifically, information entropy is a metric in information theory that measures the uncertainty of a system. Its calculation formula is as follows:

[0087] Here, X represents a random variable, and P(x) represents the probability of event x occurring. The unit of information entropy is usually bit. Information utility value = 1 - information entropy. Taking volatility tolerance as an example, assuming the information entropy of volatility tolerance is 0.72, the information utility value is 0.28. Combining the sum of the information utility values ​​of quantitative indicators (assuming it is 0.92), then the proportional coefficient of volatility tolerance is 0.28 / 0.92 = 0.30.

[0088] In one or more embodiments, acquiring fused data that affects user satisfaction includes: The system obtains performance data from the operator's system and captures user voice recordings after telephone conversations. It then analyzes and processes the voice recordings to extract voice feature data. Finally, it fuses the performance data and voice feature data to obtain fused data that influences user satisfaction.

[0089] The acquisition of performance data from the operator's system includes: acquiring performance data from the operator's system, specifically, acquiring structured service work order data from the government and enterprise operation and maintenance system, such as processing time and problem classification; and acquiring network performance data from the network management and performance system, such as packet loss rate and latency fluctuation.

[0090] Voice feature data includes voice-text data, emotional state data, and entity label data. A suitable AI voice analysis engine and deep neural network (Wave2Vec 2.0) were selected to extract voiceprint features from the call recordings. The OpenSMILE toolkit was used to identify 137 acoustic parameters, including fundamental frequency, speech rate, and pause intervals. A Bi-LSTM emotion classification model was employed to capture seven categories of emotional states in real time, such as anger (e.g., high-frequency sound intensity) and anxiety (e.g., irregular breathing sounds). Entity label data was obtained based on the core demands of the call, such as automatically labeling entities like "network latency > 500ms" and "dispute over data plan pricing." In one or more embodiments, before calculating the indicator data corresponding to each quantitative indicator based on a plurality of preset quantitative indicators and fused data, the method further includes: By utilizing graph databases and fused data, a relationship network is established, which represents the ternary relationship between users, the services they enjoy, and the networks they use. Based on the relationship network, it is confirmed that the fused data has not reached a specific threshold. The specific threshold represents the maximum probability threshold for user-initiated complaints.

[0091] Specifically, using the Neo4j graph database and fused data, a "user-service-network" relationship network is established to represent the ternary relationship between users, the services they enjoy, and the networks they use. Core entities include: User node: Basic attributes (package type / terminal device); Service node: Work order processing records (response time / resolution rate); Network node: Base station location / historical fault events; By using GraphSAGE graph neural networks to perform relationship reasoning on the relationship network, when the fused data (or relationship network) reaches a certain threshold, proactive care is immediately initiated for the user. For example, if a user experiences a delay of more than 500ms three times consecutively, it can be considered that a certain threshold has been reached. There is no need to predict the user's satisfaction anymore; proactive care and appropriate remedial measures should be taken immediately. Conversely, if the certain threshold is not reached, feature engineering is then initiated.

[0092] In one or more embodiments, it further includes: If user satisfaction is less than the preset threshold, a care strategy for the user will be generated based on the user's data under various quantitative indicators. And, based on the calculation formula, predict the user satisfaction of all users; Based on the user satisfaction of all users, target users are selected, including at least one of the following: potential churned users and high-value users.

[0093] According to another aspect of the embodiments of this disclosure, a user satisfaction prediction device is provided, such as... Figure 4 As shown, the device includes: The acquisition module 401 is used to acquire fused data that affects user satisfaction, wherein the fused data is obtained based on call voice and operator system performance data; The first calculation module 402 is used to calculate the indicator data corresponding to each quantitative indicator based on multiple preset quantitative indicators and fused data; wherein, the quantitative indicators are used to characterize the user's tolerance, scenario needs and emotional intensity. The second calculation module 403 is used to calculate the proportional coefficient of each quantitative indicator based on the fluctuation of the indicator data. Formula generation module 404 is used to obtain calculation formulas for predicting user satisfaction based on various quantitative indicators and proportional coefficients.

[0094] In one or more embodiments, the first computing module 402 is used for: The network performance data in the fused data is obtained, and the index data corresponding to the fluctuation tolerance is calculated based on the pre-trained neural network model and the network performance data. Among them, the network performance data includes packet loss rate, latency and jitter, and the fluctuation tolerance is used to predict the probability of user-initiated complaints. In addition, traffic data is acquired from the fused data, and based on the traffic data, the corresponding indicator data of the scenario sensitivity coefficient is calculated; where traffic data is the data generated when users access application software, and the scenario sensitivity coefficient is used to characterize the user's scenario needs. Furthermore, based on the fused data, identify the user's identity characteristics; based on the identity characteristics, calculate the indicator data corresponding to the service timeliness demand duration; Additionally, the system acquires speech feature data from the fused data, and calculates the corresponding index data for speech emotion recognition coefficient and text difference coefficient based on the speech feature data. The speech feature data is used to characterize the user's emotions and the intensity of their word choice.

[0095] In one or more embodiments, the second computing module 403 is used for: The indicator data is subjected to min-max normalization to obtain the normalized data corresponding to each quantitative indicator; Based on normalized data, the information entropy of each quantitative indicator is calculated; whereby information entropy is used to characterize the degree of fluctuation of normalized data. Calculate the information utility value of each quantitative indicator based on information entropy; Calculate the sum of information utility values, divide the information utility value of each quantitative indicator by the sum, and obtain the proportional coefficient of each quantitative indicator.

[0096] In one or more embodiments, the acquisition module 401 is used for: The performance data is obtained from the operator's system, including service ticket structured data and network performance data. If a phone conversation has been conducted with the user, obtain the user's voice recordings. The voice call is analyzed and processed to extract voice feature data. The voice feature data includes voice text data, emotional state data, and entity label data. The entity label data is obtained based on the core needs of the voice call. By fusing performance data and voice feature data, we can obtain fused data that influences user satisfaction.

[0097] The user satisfaction prediction device is also used to: before calculating the indicator data corresponding to each quantitative indicator based on multiple preset quantitative indicators and fused data, establish a relationship network using graph database and fused data. The relationship network is used to represent the ternary relationship between the user, the service enjoyed by the user, and the network used by the user. Based on the relationship network, it was confirmed that the fused data did not reach a specific threshold; where the specific threshold is represented by the maximum probability threshold of user-initiated complaints.

[0098] The user satisfaction prediction device is also used to: if user satisfaction is less than a preset threshold, generate a care strategy for the user based on the user's indicator data under various quantitative indicators; Furthermore, based on the calculation formula, the user satisfaction of all users is predicted; based on the user satisfaction of all users, target users are selected, wherein target users include at least one of potential churned users and high-value users.

[0099] The user satisfaction prediction device and the user satisfaction prediction method provided in this disclosure are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0100] This disclosure also provides a computer device for performing the above-described user satisfaction prediction method. Please refer to... Figure 5 It illustrates a schematic diagram of a computer device provided by some embodiments of this disclosure. For example... Figure 5 As shown, the computer device 5 includes: a processor 500, a memory 501, a bus 502, and a communication interface 503. The processor 500, the communication interface 503, and the memory 501 are connected via the bus 502. The memory 501 stores a computer program that can run on the processor 500. When the processor 500 runs the computer program, it executes the user satisfaction prediction method provided in any of the foregoing embodiments of this disclosure.

[0101] The memory 501 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0102] Bus 502 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 501 is used to store programs. After receiving an execution instruction, the processor 500 executes the program. The user satisfaction prediction method disclosed in any of the foregoing embodiments of this disclosure can be applied to the processor 500, or implemented by the processor 500.

[0103] The processor 500 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 500 or by instructions in software form. The processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPTA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 501. The processor 500 reads the information in memory 501 and, in conjunction with its hardware, completes the steps of the above method.

[0104] The computer device provided in this disclosure and the user satisfaction prediction method provided in this disclosure are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0105] This disclosure also provides a computer-readable storage medium corresponding to the user satisfaction prediction method provided in the foregoing embodiments. The computer-readable storage medium is an optical disc, on which a computer program (i.e., a computer program product) is stored. When the computer program is run by a processor, it executes the user satisfaction prediction method provided in any of the foregoing embodiments.

[0106] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0107] The computer-readable storage medium provided in the above embodiments of this disclosure and the user satisfaction prediction method provided in the embodiments of this disclosure are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0108] This disclosure also provides a computer program product; please refer to [reference needed]. Figure 6 The computer program product 600 carries program code, namely computer program 601. The instructions included in the computer program 601 can be used to execute the steps of the user satisfaction prediction method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0109] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0110] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0111] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0112] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0113] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0114] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0115] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0116] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for predicting user satisfaction, characterized in that, include: Acquire fused data that affects user satisfaction, wherein the fused data is obtained based on call voice and operator system performance data; Based on multiple preset quantitative indicators and the fused data, the indicator data corresponding to each quantitative indicator is calculated; wherein, the quantitative indicators are used to characterize the user's tolerance, scenario needs and emotional intensity. Based on the degree of fluctuation of the aforementioned indicator data, calculate the proportional coefficient of each of the aforementioned quantitative indicators; Based on the quantitative indicators and the proportional coefficients, a calculation formula for predicting user satisfaction is obtained.

2. The user satisfaction prediction method as described in claim 1, characterized in that, The quantitative indicators include fluctuation tolerance, scenario sensitivity coefficient, service timeliness requirement duration, voice emotion recognition coefficient, and text difference coefficient. Based on multiple preset quantitative indicators and the fused data, the indicator data corresponding to each quantitative indicator is calculated, including: The network performance data in the fused data is obtained, and the index data corresponding to the fluctuation tolerance is calculated based on the pre-trained neural network model and the network performance data; wherein, the network performance data includes packet loss rate, latency and jitter, and the fluctuation tolerance is used to predict the probability of the user actively complaining; In addition, traffic data is obtained from the fused data, and based on the traffic data, indicator data corresponding to the scenario sensitivity coefficient is calculated; wherein, the traffic data is the data generated when the user accesses the application software, and the scenario sensitivity coefficient is used to characterize the user's scenario needs; Furthermore, based on the fused data, the user's identity characteristics are identified; based on the identity characteristics, the indicator data corresponding to the service timeliness requirement duration is calculated; In addition, the speech feature data in the fused data is obtained, and based on the speech feature data, the index data corresponding to the speech emotion recognition coefficient and the text difference coefficient are calculated, wherein the speech feature data is used to characterize the user's emotions and the intensity of word usage.

3. The user satisfaction prediction method as described in claim 1, characterized in that, Based on the volatility of the aforementioned indicator data, the proportional coefficients of each of the quantitative indicators are calculated, including: The index data is subjected to min-max normalization to obtain the normalized data corresponding to each of the quantitative indicators; Based on the normalized data, the information entropy of each of the quantitative indicators is calculated; wherein, the information entropy is used to characterize the degree of fluctuation of the normalized data; Based on the information entropy, calculate the information utility value of each of the quantitative indicators; Calculate the sum of the information utility values, divide the information utility value of each quantitative indicator by the sum, and obtain the proportional coefficient of each quantitative indicator.

4. The user satisfaction prediction method as described in claim 1, characterized in that, Obtain integrated data that influences user satisfaction, including: The performance data is obtained from the operator's system, wherein the performance data includes service ticket structured data and network performance data; If the user has communicated with the user by phone, obtain the user's voice recording. The voice call is analyzed and processed to extract voice feature data; wherein, the voice feature data includes voice text data, emotional state data and entity label data, and the entity label data is obtained based on the core needs of the voice call. The performance data and the voice feature data are fused to obtain fused data that affects user satisfaction.

5. The user satisfaction prediction method as described in claim 1, characterized in that, Before calculating the indicator data corresponding to each of the preset quantitative indicators and the fused data, the method further includes: Using the graph database and the fused data, a relationship network is established, which is used to represent the ternary relationship between the user, the services enjoyed by the user, and the network used by the user. Based on the relationship network, it is confirmed that the fused data has not reached a specific threshold; wherein, the specific threshold is characterized as the maximum probability threshold of the user's proactive complaint.

6. The user satisfaction prediction method as described in claim 1, characterized in that, Also includes: If the user satisfaction is less than a preset threshold, a care strategy for the user is generated based on the user's indicator data under each of the quantitative indicators. And, based on the calculation formula, predict the user satisfaction of all users; Based on the user satisfaction of all users, target users are selected, wherein the target users include at least one of potential churned users and high-value users.

7. A user satisfaction prediction device, characterized in that, include: The acquisition module is used to acquire fused data that affects user satisfaction, wherein the fused data is obtained based on call voice and operator system performance data; The first calculation module is used to calculate the indicator data corresponding to each of the preset quantitative indicators and the fused data; wherein, the quantitative indicators are used to characterize the user's tolerance, scenario needs and emotional intensity. The second calculation module is used to calculate the proportional coefficient of each of the quantitative indicators based on the fluctuation level of the indicator data. The formula generation module is used to obtain a calculation formula for predicting user satisfaction based on the quantitative indicators and the proportional coefficients.

8. A computer embedded device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 6.

Images