A dynamic resource allocation method and system for video customer service analysis tasks

Abstract

The present invention discloses a dynamic resource allocation method and system for a video customer service analysis task, and relates to the technical field of dynamic resource allocation. A trained customer service rating model is used to rate customer service, and when receiving customers, collected consultation reception data is aggregated to generate a customer consultation data set; risk customers are identified among the received customers, and the concentration of risk customers is constructed according to the time node when the risk customers are identified. If the concentration exceeds the concentration threshold, a customer service resource scheduling strategy is generated using a trained customer service resource scheduling model, and customer service of an adaptive level is allocated to the risk customer among several customer services whose number of pending requests is lower than expected; a feedback value is constructed after recording the risk customer churn rate and satisfaction value, and corresponding dynamic adjustment measures are taken for the allocation of customer resources according to the feedback value; and the customer churn risk is reduced by realizing dynamic optimization and adjustment of the customer resource scheduling strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic resource allocation, and in particular to a dynamic resource allocation method and system for a video customer service analysis task. Background Art

[0002] Video customer service is a high-tech means of using video streaming technology to enable instant communication between customers and customer service personnel through video. It uses web pages or video equipment as carriers and the latest network technology to provide website visitors or customers with an intuitive and vivid communication experience. Compared with traditional text or voice customer service, video customer service has a higher degree of visualization and richer information transmission content, which can significantly improve communication effectiveness and customer satisfaction.

[0003] Video customer service resource allocation refers to the reasonable allocation and arrangement of the working hours and tasks of video customer service personnel based on factors such as customer needs, the capabilities of the customer service team, and business peak hours, so as to ensure efficient and timely response and handling of customer inquiries and problems. This process aims to optimize customer experience and improve customer satisfaction, while ensuring the maximum efficiency and resource utilization of the customer service team. Through intelligent scheduling systems and data analysis, companies can accurately predict customer needs, flexibly adjust customer service resources, and achieve optimal resource allocation.

[0004] In the customer service resource allocation method, device, storage medium and computer equipment provided by the present invention with application publication number CN114140016A, when receiving the manual customer service request instruction triggered by the user, the entry request parameters of the business involved when the user triggers the manual customer service request instruction can be first determined, and then the connection queue matching the entry request parameters can be determined. Since the manual customer service in the connection queue matching the entry request parameters is mainly used to handle problems related to the business corresponding to the entry request parameters, after determining the connection queue, the manual customer service in the connection queue can be selected as the object for online communication with the user; in this way, the manual customer service in different connection queues only needs to be trained on the business knowledge corresponding to the connection queue before taking up the post, and does not need to master the business knowledge of the entire node, thereby greatly shortening the training cycle and adaptation process of the manual customer service, and reducing the investment cost of manual resources to a certain extent.

[0005] Combining the above content with the content in the prior art:

[0006] When the customer service access system is in working condition, after the customer accesses, if the auxiliary intelligent voice cannot solve the customer's consultation problem, it is usually necessary to assign professional customer service to the customer, who will receive the customer and answer and handle the consultation content. However, considering that when the customer service access system allocates customer service resources, it usually randomly assigns customer service to customers, and there may be large differences between different customer service skill reserves and professional capabilities. Therefore, some customers' complex consultation content may be difficult to be effectively resolved, and their satisfaction will be poor; in order to solve this problem, the existing customer service access system will consider the customer's historical access situation after accessing the customer, so as to push and match the customer with better professional capabilities to ensure customer satisfaction.

[0007] However, when a large number of customers are connected at the same time, especially when there are large differences in the consultation content of different customers, and the customer service satisfaction feedback in the current stage is also poor, the customer service will be under greater pressure when receiving customer inquiries; although the existing customer service access system has the ability to push customer service, it is insufficient in the scheduling and allocation of customer service resources in complex scenarios. When the satisfaction is low, some customers with potential loss risks may still be lost.

[0008] To this end, the present invention provides a method and system for dynamic resource allocation of video customer service analysis tasks. Summary of the invention

[0009] 1. Technical issues to be resolved

[0010] In view of the deficiencies in the prior art, the present invention provides a dynamic resource allocation method and system for video customer service analysis tasks, which generates a customer consultation data set by aggregating the collected consultation reception data; identifying risky customers among the received customers, and building a concentration of risky customers based on the time nodes when the risky customers are identified; if the concentration exceeds the concentration threshold, using the trained customer service resource scheduling model to generate a customer service resource scheduling strategy, and among several customer services whose number of pending requests is lower than expected, allocating customer services of an adaptation level to the risky customers; recording the risky customer churn rate and satisfaction value and then building a feedback value, and taking corresponding dynamic adjustment measures for the allocation of customer resources based on the feedback value; reducing the risk of customer churn by realizing dynamic optimization and adjustment of the customer resource scheduling strategy, thereby solving the technical problems recorded in the background technology.

[0011] (II) Technical solution

[0012] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0013] A dynamic resource allocation method and system for video customer service analysis tasks, including analyzing the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued;

[0014] Taking the customer service reception data as input, the trained customer service rating model is used to rate the customer service, and when receiving the customer, the collected consultation reception data is aggregated to generate a customer consultation data set;

[0015] Use the trained churn risk identification model to identify risky customers among the customers received, and build the concentration of risky customers based on the time nodes when risky customers were identified , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world;

[0016] After identifying the complexity of the customer's consultation content, the trained customer service resource scheduling model is used to generate a customer service resource scheduling strategy. Among several customer service centers where the number of pending requests is lower than expected, customer service centers with appropriate levels are assigned to risky customers.

[0017] Record the risk customer churn rate and satisfaction value and build feedback value and feedback threshold , according to the feedback value With feedback threshold The relationship between customer resources and the churn rate is analyzed and corresponding dynamic adjustment measures are taken for the allocation of customer resources. and satisfaction After dimensionless processing, the following is done:

[0018]

[0019] Weight coefficient, , , is a constant correction factor.

[0020] Furthermore, the customer's consultation content data is recorded, and similarity analysis is performed on different consultation content data. The difference corresponding to the consultation content is obtained by obtaining similarity data analysis. , when the difference in customer inquiries When the preset difference threshold is exceeded, a satisfaction analysis instruction is issued to the outside.

[0021] Furthermore, the similarity between the two consultation contents Perform linear normalization and map the corresponding data values ​​to the interval The difference is constructed according to the following formula :

[0022]

[0023] Weight coefficient, , ,and , is the similarity between the i-th consulting content and the j-th consulting content, is the mean similarity value, , is the number of consulting contents; the weight coefficient can be obtained by referring to the hierarchical analysis method;

[0024] Furthermore, after receiving the satisfaction analysis instruction, the customer's satisfaction data within the collection period is collected; the customer's satisfaction value is constructed from the satisfaction data , among which, in satisfaction Under dimensionless conditions, according to the following formula:

[0025]

[0026] in, is the satisfaction of the i-th customer, is the mean value of satisfaction, is the weight coefficient, .

[0027] Furthermore, after receiving the rating instruction, customer service reception data is collected, and part of the data in the customer service reception data set is extracted to train the convolutional neural network to obtain the trained customer service rating model; during the data collection cycle, after the customer ends the reception, the consultation reception data is recorded and collected, and summarized to generate a customer consultation data set.

[0028] Further, the lost customer information and the corresponding consultation reception data within the data collection period are collected, and features are extracted from the corresponding data to obtain the lost customer's lost features;

[0029] Using customer consultation reception data as input, the trained churn risk identification model identifies the customer's churn risk. If the customer's churn risk exceeds expectations, it will be treated as a risky customer.

[0030] Furthermore, after a number of risky customers appear in succession, the concentration of risky customers is constructed based on the time nodes at which the risky customers are identified. , as follows:

[0031]

[0032] Weight coefficient, , ; n is the number of risky customers, is the time interval from the i-th appearance of a risky customer to the j-th appearance of a risky customer, is the average value of the time interval.

[0033] Furthermore, after receiving the early warning instruction, if a risky customer is accessed again, the customer's consultation data is used as input, and the complexity of the customer's consultation content is identified by the trained consultation content complexity identification model;

[0034] Taking reducing the churn rate of risky customers as the scheduling goal and not increasing the customer service reception load as the constraint condition, after determining the complexity of the risky customer consultation content and the historical satisfaction data, the trained customer service resource scheduling model generates a customer service resource scheduling strategy.

[0035] Furthermore, after obtaining several feedback values ​​continuously, a feedback threshold is constructed. , as follows:

[0036] in, , k is the number of sub-periods, is the feedback value in the ith sub-cycle, is the mean of the feedback values.

[0037] Further, based on the feedback value With feedback threshold The relationship between the two is to adopt corresponding adjustment strategies for customer access. Exceeding feedback threshold When the feedback value is At the feedback threshold After collecting reception feedback data, we use the multi-objective algorithm to optimize the customer service resource scheduling strategy, take improving the feedback value as the optimization goal, obtain the optimized customer service resource scheduling strategy and execute it; when the feedback value Below feedback threshold When the risk customers are at risk, a follow-up visit stage is added.

[0038] A dynamic resource allocation method and system for video customer service analysis tasks, including a difference analysis unit, which analyzes the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued;

[0039] The consultation data collection unit takes the customer service reception data as input, uses the trained customer service rating model to rate the customer service, and aggregates the collected consultation reception data to generate a customer consultation data set when receiving the customer;

[0040] The churn risk unit uses the trained churn risk identification model to identify risky customers among the customers received, and builds the concentration of risky customers based on the time nodes when risky customers are identified. , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world;

[0041] The resource scheduling unit, after identifying the complexity of the customer consultation content, uses the trained customer service resource scheduling model to generate a customer service resource scheduling strategy, and assigns customer service with an adaptive level to risky customers among several customer service staffs whose number of pending requests is lower than expected;

[0042] Feedback processing unit, records the risk customer churn rate and satisfaction value and then builds the feedback value and feedback threshold , according to the feedback value With feedback threshold The relationship between customers and enterprises is determined, and corresponding dynamic adjustment measures are taken for the allocation of customer resources.

[0043] (III) Beneficial effects

[0044] The present invention provides a method and system for dynamic resource allocation of video customer service analysis tasks, which has the following beneficial effects:

[0045] 1. Based on the similarity and complexity of different problems, targeted supplementation and training for customer service staff can improve customer satisfaction; based on satisfaction value Customer satisfaction feedback can be evaluated as a whole. If customer satisfaction is insufficient at the current stage, timely adjustments can be made.

[0046] 2. By determining the level of customer service, customers who are more difficult to serve will be served by customer service staff with stronger capabilities and higher levels, thereby improving customer satisfaction. By screening and identifying customers with loss risks among a number of customers, targeted reception strategies or marketing strategies can be formulated to reduce customer loss rates.

[0047] 3. Based on the degree of aggregation The customer reception effect in the previous stage can be evaluated and fed back, and this can be used as a feedback indicator to determine the concentration of risky customers. When it is higher or above expectations, the current customer reception strategy can be adjusted and optimized to reduce customer churn rate.

[0048] 4. The trained customer service resource scheduling model schedules online customer service staff and allocates customer service resources to avoid waste of customer service resources and reduce the churn rate of risky customers.

[0049] 5. Based on feedback value The effectiveness of executing customer service resource scheduling strategies can be evaluated, and when the customer churn problem cannot be improved, customer service resources can be reallocated in a timely manner.

[0050] 6. Based on the relationship between the feedback value and the feedback threshold, customer consultation reception is put in a state of dynamic feedback and dynamic evaluation, so that the current reception feedback evaluation has better accuracy and pertinence.

[0051] 7. Based on feedback value With feedback threshold By adopting different handling strategies based on the relationship between customers and achieving mutual coordination of different measures when reducing customer churn, better results can be achieved; by dynamically optimizing and adjusting the customer resource scheduling strategy, the customer service resource allocation and risk customer access are always in a matched state, thereby minimizing the customer churn risk; and further, for customers with a higher risk of churn, a return visit process is added to achieve secondary processing of the consulting content, supplement the deficiencies in customer service resource scheduling, and reduce customer churn rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic diagram of a flow chart of a dynamic resource allocation method for a video customer service analysis task of the present invention;

[0053] Figure 2 A schematic diagram of the structure of a dynamic resource allocation system for video customer service analysis tasks of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] See also Figure 1 The present invention provides a dynamic resource allocation method for a video customer service analysis task, comprising:

[0056] Step 1: Analyze the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued;

[0057] The step 1 includes the following contents:

[0058] Step 101: Record the customer's consultation content data within a preset data collection period, perform similarity analysis on different consultation content data, and obtain the difference degree corresponding to the consultation content from the similarity data analysis. , where the similarity between the two consulting contents is Perform linear normalization and map the corresponding data values ​​to the interval The difference is constructed according to the following formula :

[0059]

[0060] Weight coefficient, , ,and , is the similarity between the i-th consulting content and the j-th consulting content, is the mean similarity value, , is the number of consulting contents; the weight coefficient can be obtained by referring to the hierarchical analysis method;

[0061] When customers inquire about the difference When the difference exceeds the preset threshold, it is necessary to count the customer's satisfaction with the customer service reception after completing the consultation. At this time, a satisfaction analysis instruction is issued to the outside.

[0062] When in use, after continuously collecting consultation content data of several groups of customers, by conducting similarity analysis on these consultation contents, targeted supplementation and training can be provided to customer service staff based on the similarity and complexity of different questions, thus improving customer satisfaction;

[0063] Step 102: after receiving the satisfaction analysis instruction, conduct a satisfaction survey on the customer after the customer finishes the consultation, collect the feedback questionnaire sent to the customer, and collect the customer's satisfaction data within the collection period;

[0064] Construct the customer satisfaction value in the current stage based on satisfaction data , among which, in satisfaction Under dimensionless conditions, according to the following formula:

[0065]

[0066] in, is the satisfaction of the i-th customer, is the mean value of satisfaction, is the weight coefficient, ; The number of customers;

[0067] Based on historical data and management expectations of customer service reception feedback, set the satisfaction threshold in advance; if the satisfaction value obtained is If the satisfaction level is lower than the satisfaction threshold, it means that during the current data collection period, the customer satisfaction level after completing the consultation is relatively low. If it is not handled in time, there may be a risk of customer loss. At this time, a rating instruction is issued to the outside.

[0068] When using, combine the contents in steps 101 and 102:

[0069] After obtaining satisfaction data from several customers, build a satisfaction value based on this , based on satisfaction value Customer satisfaction feedback can be evaluated as a whole. If customer satisfaction is insufficient at the current stage, it means that the feedback from the current customer reception process is poor and needs to be adjusted in a timely manner, otherwise there will be a risk of customer loss.

[0070] When a large number of customers are connected at the same time, especially when there are large differences in the consultation content of different customers, and the customer service satisfaction feedback in the current stage is also poor, the customer service will be under greater pressure when receiving customer inquiries; although the existing customer service access system has the ability to push customer service, it is insufficient in the scheduling and allocation of customer service resources in complex scenarios. When the satisfaction is low, some customers with potential churn risk may still churn.

[0071] Step 2: Using the customer service reception data as input, use the trained customer service rating model to rate the customer service, and when receiving the customer, aggregate the collected consultation reception data to generate a customer consultation data set;

[0072] The step 2 includes the following contents:

[0073] Step 201: after receiving the rating instruction, customer service reception data is collected, such as customer reception and communication data, customer consultation data, average customer waiting time, customer satisfaction, one-time resolution rate and service failure rate, etc., and summarized as a customer service reception data set;

[0074] Extract some data from the customer service reception data set, annotate it and use it as sample data; train the convolutional neural network with the annotated sample data to obtain the trained customer service rating model; use the customer service reception data as input, and use the trained customer service rating model to rate the customer service, and mark the customer service with the obtained rating;

[0075] When in use, with the assistance of historical data, the trained customer service rating model rates customers. By determining the level of customer service, customers who are more difficult to serve, such as customers with a greater risk of churn, are served by customer service staff with stronger capabilities and higher levels, thus ensuring the effectiveness of customer service and improving customer satisfaction.

[0076] Step 202: During the data collection period, after the customer has finished the consultation, the content and type, processing time, resolution results, customer feedback data, etc. of each consultation, as well as common complex issues, such as technical support requests, account issues or complaints, are recorded and collected, and the feedback data provided by the customer is recorded; after the above data are used as consultation reception data, a customer consultation data set is generated;

[0077] When using, combine the contents in steps 201 and 202:

[0078] After the customer finishes the reception process, by collecting and recording various consultation feedback data, it can serve as feedback and reference when classifying and labeling customers.

[0079] Step 3: Use the trained churn risk identification model to identify risky customers among the customers received, and build the concentration of risky customers based on the time nodes when risky customers were identified. , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world;

[0080] The step three includes the following contents:

[0081] Step 301: collect information on lost customers and corresponding consultation reception data within a data collection period, extract features from the corresponding data, and obtain lost features of lost customers, such as consultation type, resolution time, etc.;

[0082] After training the random forest algorithm with the labeled sample data, a trained churn risk identification model is obtained; using the customer consultation reception data as input, the trained churn risk identification model identifies the customer's churn risk. If the customer's churn risk exceeds expectations, it is regarded as a risk customer and the identified risk customer is marked;

[0083] When in use, based on the historical data collected, the trained churn risk identification model identifies the churn risk of customers, and screens and identifies customers with churn risk among a number of customers. For such risky customers, targeted reception strategies or marketing strategies can be formulated to reduce customer churn rate;

[0084] Step 302: After a number of risky customers appear continuously in the current stage, under dimensionless conditions, the concentration of risky customers is constructed based on the time nodes at which the risky customers are identified. , as follows:

[0085]

[0086] Weight coefficient, , ; The weight coefficient can be obtained by referring to the hierarchical analysis method; n is the number of risky customers, is the time interval from the i-th appearance of a risky customer to the j-th appearance of a risky customer, is the time interval average;

[0087] Pre-set aggregation thresholds based on historical data and expectations for customer churn risk management;

[0088] If the concentration If the concentration threshold is exceeded, it means that the concentration of risky customers is high in the current stage. If the customer service fails to respond effectively during reception, a large number of risky customers may be lost. At this time, an early warning instruction is issued to the outside world;

[0089] When using, combine the contents in steps 301 and 302:

[0090] After connecting to several customers with churn risk, analyze the concentration of risky customers , according to the concentration The customer reception effect in the previous stage can be evaluated and fed back, and this can be used as a feedback indicator to determine the concentration of risky customers. When it is higher or above expectations, the current customer reception strategy can be adjusted and optimized to avoid risky customers from turning into lost customers, thereby reducing customer churn rate.

[0091] Step 4: After identifying the complexity of the customer consultation content, use the trained customer service resource scheduling model to generate a customer service resource scheduling strategy, and assign customer service with an adaptive level to risky customers among several customer service centers where the number of pending requests is lower than expected;

[0092] The step 4 includes the following contents:

[0093] Step 401: After receiving the early warning instruction, the sample data recurrent neural network is used to obtain a trained consultation content complexity recognition model; if a risky customer is accessed again, the consultation data of the risky customer is recorded, and the consultation data of the customer is used as input. The complexity of the customer's consultation content is identified by the trained consultation content complexity recognition model, and the accessed customer is marked with the complexity;

[0094] Step 402: train the ant colony algorithm with the sample data to obtain a trained customer service resource scheduling model; take reducing the churn rate of risky customers as the scheduling target, take not increasing the customer service reception load as the constraint condition, and after determining the complexity of the risky customer consultation content and historical satisfaction data, generate a customer service resource scheduling strategy with the trained customer service resource scheduling model;

[0095] Execute customer service resource scheduling strategies, and assign customer service staff of appropriate levels to risky customers from the customer service pool that is currently online and has a lower-than-expected number of pending requests;

[0096] When using, combine the contents in steps 401 and 402:

[0097] When risky customers are contacted again, the trained customer service resource scheduling model will schedule online customer service personnel based on the complexity of the consultation content and the customer service level, and allocate customer service resources, so that customer service personnel with better personal abilities and better training can serve customers with higher churn risks and more complex consultation content that is difficult to resolve, thus avoiding waste of customer service resources and reducing the churn rate of risky customers.

[0098] Step 5: Record the risk customer churn rate and satisfaction value and then build the feedback value and feedback threshold , according to the feedback value With feedback threshold and take corresponding dynamic adjustment measures for the allocation of customer resources;

[0099] The step five includes the following contents:

[0100] Step 501: pre-set an observation period including several sub-periods, record the risk customer churn rate and satisfaction value in the sub-periods, and then construct a feedback value , where the churn rate and satisfaction After dimensionless processing, refer to the hierarchical analysis method to determine the corresponding weight coefficient, and then follow the following method:

[0101]

[0102] The weight coefficient is, , , is a constant correction coefficient whose value falls within ;

[0103] When in use, feedback values ​​are generated by recording customer churn data and satisfaction data in each sub-stage , according to the feedback value The effectiveness of executing customer service resource scheduling strategies can be evaluated. When the customer churn problem cannot be improved, customer service resources can be reallocated in a timely manner.

[0104] Construct a feedback threshold after obtaining several feedback values ​​in succession , as follows:

[0105] in, , k is the number of sub-periods, is the feedback value in the ith sub-cycle, is the mean of the feedback values;

[0106] When in use, by building a dynamic feedback threshold based on the feedback value, and based on the relationship between the feedback value and the feedback threshold, the customer consultation reception can be put in a state of dynamic feedback and dynamic evaluation, so that the current reception feedback evaluation has better accuracy and pertinence;

[0107] Step 502: Based on the feedback value With feedback threshold The relationship between the two is to adopt corresponding adjustment strategies for customer access. Exceeding feedback threshold When the current customer service resource scheduling strategy is continued;

[0108] When the feedback value At the feedback threshold During the process, after collecting reception feedback data, such as customer churn rate, satisfaction data and feedback, the customer service resource scheduling strategy is optimized using a multi-objective algorithm with the improvement of feedback value as the optimization goal, and the optimized customer service resource scheduling strategy is obtained and executed;

[0109] When the feedback value Below feedback threshold When the risk customers are at risk, a return visit phase is added, and the risk customers are returned during the return visit phase;

[0110] When using, combine the contents in steps 501 and 502:

[0111] Based on feedback value With feedback threshold By adopting different handling strategies based on the relationship between customers and risks, different measures can be coordinated to achieve better results when reducing customer churn. Moreover, by dynamically optimizing and adjusting the customer resource scheduling strategy, customer service resource allocation and risk customer access are always in a matching state, thereby minimizing the customer churn risk. Furthermore, for customers with a higher risk of churn, a return visit process is added to achieve secondary processing of the consulting content and supplement the deficiencies in customer service resource scheduling, thereby reducing customer churn rate.

[0112] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0113] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0114] See also Figure 2 The present invention provides a dynamic resource allocation system for video customer service analysis tasks, comprising:

[0115] Difference analysis unit, analyzing the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued;

[0116] The consultation data collection unit takes the customer service reception data as input, uses the trained customer service rating model to rate the customer service, and aggregates the collected consultation reception data to generate a customer consultation data set when receiving the customer;

[0117] The churn risk unit uses the trained churn risk identification model to identify risky customers among the customers received, and builds the concentration of risky customers based on the time nodes when risky customers are identified. , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world;

[0118] The resource scheduling unit, after identifying the complexity of the customer consultation content, uses the trained customer service resource scheduling model to generate a customer service resource scheduling strategy, and assigns customer service with an adaptive level to risky customers among several customer service staffs whose number of pending requests is lower than expected;

[0119] Feedback processing unit, records the risk customer churn rate and satisfaction value and then builds the feedback value and feedback threshold , according to the feedback value With feedback threshold The relationship between customers and enterprises is determined, and corresponding dynamic adjustment measures are taken for the allocation of customer resources.

[0120] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0121] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0122] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0123] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as 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 devices or units, which can be electrical, mechanical or other forms.

[0124] 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 network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0125] 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.

[0126] If the function 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 can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or 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, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0127] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A dynamic resource allocation method for a video customer service analysis task, characterized in that: include, Analyze the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued; Taking the customer service reception data as input, the trained customer service rating model is used to rate the customer service. When receiving customers, the collected consultation reception data is aggregated to generate a customer consultation data set. Use the trained churn risk identification model to identify risky customers among the customers received, and build the concentration of risky customers based on the time nodes when risky customers were identified , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world; After identifying the complexity of the customer's consultation content, the trained customer service resource scheduling model is used to generate a customer service resource scheduling strategy. Among several customer service centers where the number of pending requests is lower than expected, customer service centers with appropriate levels are assigned to risky customers. Record the risk customer churn rate and satisfaction value and build feedback value and feedback threshold , according to the feedback value With feedback threshold The relationship between customer resources and the churn rate is analyzed and corresponding dynamic adjustment measures are taken for the allocation of customer resources. and satisfaction After dimensionless processing, follow the following method: ; Weight coefficient, , , is a constant correction factor.

2. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 1, characterized in that: Record the customer's consultation content data, perform similarity analysis on different consultation content data, and obtain the difference corresponding to the consultation content through similarity data analysis ; When the customer inquires about the difference When the preset difference threshold is exceeded, a satisfaction analysis instruction is issued to the outside.

3. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 2, characterized in that: After receiving the satisfaction analysis instruction, collect the customer's satisfaction data within the collection period; construct the customer's satisfaction value based on the satisfaction data , among which, in satisfaction Under dimensionless conditions, according to the following formula: ;in, is the satisfaction of the i-th customer, is the mean value of satisfaction, is the weight coefficient, .

4. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 3, characterized in that: After receiving the rating instruction, the customer service reception data is collected, and part of the data in the customer service reception data set is extracted to train the convolutional neural network to obtain the trained customer service rating model; during the data collection cycle, after the customer ends the reception, the consultation reception data is recorded and collected, and summarized to generate a customer consultation data set.

5. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 4, characterized in that: Collect the lost customer information and corresponding consultation reception data during the data collection period, extract features from the corresponding data and obtain the lost customer's lost features; Using customer consultation reception data as input, the trained churn risk identification model identifies the customer's churn risk. If the customer's churn risk exceeds expectations, it will be treated as a risky customer.

6. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 5, characterized in that: After several risky customers appear in succession, the concentration of risky customers is constructed based on the time nodes at which risky customers are identified. , as follows: ; Weight coefficient, , ; n is the number of risky customers, is the time interval from the i-th appearance of a risky customer to the j-th appearance of a risky customer, is the average value of the time interval.

7. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 6, characterized in that: After receiving the early warning instruction, if a risky customer is accessed again, the customer's consultation data is used as input, and the complexity of the customer's consultation content is identified by the trained consultation content complexity recognition model; Taking reducing the churn rate of risky customers as the scheduling goal and not increasing the customer service reception load as the constraint condition, after determining the complexity of the risky customer consultation content and the historical satisfaction data, the trained customer service resource scheduling model generates a customer service resource scheduling strategy.

8. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 7, characterized in that: Construct a feedback threshold after obtaining several feedback values ​​in succession , as follows: ;in, , k is the number of sub-periods, is the feedback value in the ith sub-cycle, is the mean of the feedback values.

9. A method for dynamic resource allocation of a video customer service analysis task as claimed in claim 8, characterized in that: Based on feedback value With feedback threshold The relationship between the two is to adopt corresponding adjustment strategies for customer access, among which, When the feedback value Exceeding feedback threshold When the current customer service resource scheduling strategy is continued; When the feedback value At the feedback threshold During the process, after collecting reception feedback data, the customer service resource scheduling strategy is optimized using a multi-objective algorithm with the improvement of feedback value as the optimization goal, and the optimized customer service resource scheduling strategy is obtained and executed; When the feedback value Below feedback threshold When the risk customers are at risk, a follow-up visit stage is added.

10. A dynamic resource allocation system for video customer service analysis tasks, characterized in that: include, Difference analysis unit, analyzing the differences between customer consultation contents , when the difference When the customer's satisfaction exceeds expectations, we collect satisfaction data after the customer completes the consultation and use the satisfaction data to build the customer's satisfaction value. If the satisfaction value If it is below the satisfactory threshold, an external rating instruction will be issued; The consultation data collection unit takes the customer service reception data as input, uses the trained customer service rating model to rate the customer service, and aggregates the collected consultation reception data to generate a customer consultation data set when the customer is received; The churn risk unit uses the trained churn risk identification model to identify risky customers among the customers received, and builds the concentration of risky customers based on the time nodes when risky customers are identified. , if the aggregation If the aggregation threshold is exceeded, an early warning command will be issued to the outside world; The resource scheduling unit, after identifying the complexity of the customer consultation content, uses the trained customer service resource scheduling model to generate a customer service resource scheduling strategy, and assigns customer service with an adaptive level to risky customers among several customer service staffs whose number of pending requests is lower than expected; Feedback processing unit, records the risk customer churn rate and satisfaction value and then builds the feedback value and feedback threshold , according to the feedback value With feedback threshold The relationship between customers and enterprises is determined, and corresponding dynamic adjustment measures are taken for the allocation of customer resources.

Images